Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Survival and growth curve analyses applied to a barnacle data set Freund, Vivien 1983

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-UBC_1984_A6_7 F74.pdf [ 4.8MB ]
Metadata
JSON: 831-1.0096064.json
JSON-LD: 831-1.0096064-ld.json
RDF/XML (Pretty): 831-1.0096064-rdf.xml
RDF/JSON: 831-1.0096064-rdf.json
Turtle: 831-1.0096064-turtle.txt
N-Triples: 831-1.0096064-rdf-ntriples.txt
Original Record: 831-1.0096064-source.json
Full Text
831-1.0096064-fulltext.txt
Citation
831-1.0096064.ris

Full Text

SURVIVAL AND GROWTH CURVE ANALYSES APPLIED TO A BARNACLE DATA SET By VIVIEN FREUND B . A . , The U n i v e r s i t y o f Durham, 1981 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE i n THE FACULTY OF GRADUATE STUDIES ( I n s t i t u t e of A p p l i e d M a t h e m a t i c s and S t a t i s t i c s ) We a c c e p t t h i s t h e s i s as c o n f o r m i n g t o the r e q u i r e d s t a n d a r d THE UNIVERSITY OF BRITISH COLUMBIA December 1983 © V i v i e n F r e u n d , 1983 I n p r e s e n t i n g t h i s t h e s i s i n p a r t i a l f u l f i l m e n t o f t h e r e q u i r e m e n t s f o r an advanced degree a t t h e U n i v e r s i t y o f B r i t i s h C o l u m b i a , I agree t h a t t h e L i b r a r y s h a l l make i t f r e e l y a v a i l a b l e f o r r e f e r e n c e and s t u d y . I f u r t h e r agree t h a t p e r m i s s i o n f o r e x t e n s i v e c o p y i n g o f t h i s t h e s i s f o r s c h o l a r l y p u r p o s e s may be g r a n t e d by t h e head o f my department o r by h i s o r h e r r e p r e s e n t a t i v e s . I t i s u n d e r s t o o d t h a t c o p y i n g o r p u b l i c a t i o n o f t h i s t h e s i s f o r f i n a n c i a l g a i n s h a l l n o t be a l l o w e d w i t h o u t my w r i t t e n p e r m i s s i o n . Department o f The U n i v e r s i t y o f B r i t i s h C olumbia 1956 Main Mall V a n c ouver, Canada V6T 1Y3 Date DE-6 (3/81) ABSTRACT A d a t a s e t r e l a t i n g t o the s u r v i v a l and growth of b a r n a c l e s i s e x a m i n e d . S u r v i v a l d i s t r i b u t i o n s a r e compared by means of t h r e e n o n p a r a m e t r i c t e s t s . The e x p o n e n t i a l model i s t h e n f i t t e d f o r the s u r v i v a l d i s t r i b u t i o n s and a random e f f e c t s model i s d e v e l o p e d f o r the s l o p e . P o l y n o m i a l growth c u r v e s a r e f i t t e d and v a r i o u s h y p o t h e s e s r e l a t i n g t o the p a r a m e t e r s a r e t e s t e d u s i n g f i r s t l y the model of P o t h o f f and Roy and s e c o n d l y the model of Rao . Owing to the n a t u r e of t h e g rowth d a t a , w h i c h i s no t l o n g i t u d i n a l , t h i s p r e s e n t s v a r i o u s s t a t i s t i c a l p rob lems w h i c h a r e d i s c u s s e d . A/j John P e t k a u T h e s i s S u p e r v i s o r - i i i -TABLE OF CONTENTS Page ABSTRACT i i TABLE OF CONTENTS i i i LIST OF TABLES i v LIST OF FIGURES v ACNOWLEDGEMENTS v i 1. INTRODUCTION 1 1.1 The E x p e r i m e n t 3 1.2 The D a t a 5 2. EXAMINATION OF THE SURVIVAL DATA 6 2.1 D e s c r i p t i o n of the S u r v i v a l D a t a 8 2.2 E x p o n e n t i a l Mode l s 19 2.3 I n f e r e n c e on the B 's 31 2.4 Random E f f e c t s M o d e l 36 2.5 I n c o r p o r a t i n g N e t t i n g and D o w l i n g L i n e s i n t o the M o d e l . . 46 3. EXAMINATION OF THE GROWTH DATA 52 3.1 Growth D a t a 52 3.2 Growth Curves f o r Items 54 3 .3 The Growth Curve M o d e l s 56 3.4 A p p l i c a t i o n of the Mode l s 64 3.5 System One O y s t e r L i n e s 67 3 .6 System Two O y s t e r L i n e s 76 3.7 A p p l i c a t i o n of the M o d e l t o a l l S e v e n t e e n L i n e s 83 3 .8 E x t e n s i o n of the M o d e l t o A l l o w D i f f e r e n t C o v a r i a n c e M a t r i c e s f o r D i f f e r e n t I tems 91 3 .9 C o l l a p s e Over L i n e s W i t h i n each S y s t e m - t y p e C o m b i n a t i o n . 95 3 .10 Misusing °ata 98 3.11 Growth C u r v e s f o r I n d i v i d u a l B a r n a c l e s 101 3.12 D i s c u s s i o n 107 CONCLUSION 114 BIBLIOGRAPHY 116 - i v -L I S T OF TABLES Page T a b l e 1 S u r v i v a l Da ta 7 T a b l e 2 R e s u l t s of F i t t i n g the E x p o n e n t i a l Mode l S t a r t i n g a t t = 0 23 T a b l e 3 R e s u l t s of F i t t i n g the E x p o n e n t i a l M o d e l S t a r t i n g a t t = 5 25 - v -L I S T OF FIGURES Page F i g u r e 1 Growth Curves f o r the System One O y s t e r L i n e s 73 F i g u r e 2 Growth C u r v e s f o r the System Two O y s t e r L i n e s 73 F i g u r e 3 Growth Curves f o r the Dowl ing L i n e s 87 F i g u r e 4 Growth Curves f o r the N e t t i n g L i n e s 87 - v i -ACKNOWLEDGEMENTS I would l i k e to e x p r e s s my g r a t i t u d e to D r . John P e t k a u f o r a l l t h e t ime he spent i n h e l p i n g t o p r o d u c e t h i s t h e s i s . I am i n d e b t e d to D r . Nancy R e i d f o r h e r sympathy and encouragement a l o n g the way, and f o r r e a d i n g the t h e s i s . Many t h a n k s a l s o to He lene C repeau who a lways had t ime t o h e l p . The f i n a n c i a l s u p p o r t of the C a n a d i a n Commonwealth S c h o l a r s h i p and F e l l o w s h i p Committee i s g r a t e f u l l y a c k n o w l e d g e d . 1. INTRODUCTION The e x a m i n a t i o n of a p a r t i c u l a r d a t a s e t l e a d s , d u r i n g the c o u r s e i o f t h i s t h e s i s , t o the d i s c u s s i o n and a p p l i c a t i o n of a number o f s t a t i s t i c a l t e c h n i q u e s . These t e c h n i q u e s f a l l under the two g e n e r a l h e a d i n g s o f s u r v i v a l a n a l y s i s and growth c u r v e s . The d a t a s e t w h i c h was i n v e s t i g a t e d c o n s i s t e d of s u r v i v a l and growth d a t a f o r b a r n a c l e s . B a r n a c l e s a r e l i t t l e c r e a t u r e s which l i v e i n the o c e a n . The f i r s t p a r t of t h e i r l i v e s i s spent l o o k i n g f o r a p l a c e to l i v e , w h i c h w i l l be some o b j e c t i n the ocean such as a r o c k , a boat or a w h a l e . They u s u a l l y choose a p l a c e where t h e r e are o t h e r b a r n a c l e s . I t i s i m p o r t a n t t h a t they choose a p l a c e w i t h s t r o n g c u r r e n t s to b r i n g i n p l a n k t o n f o r f o o d . Once i t has chosen a r o c k , or some o t h e r o b j e c t , the b a r n a c l e a t t a c h e s i t s e l f f i r s t u s i n g a r e l a t i v e l y m i l d g l u e but s u b s e q u e n t l y u s i n g a d u l t g l u e w h i c h i s e x t r e m e l y s t r o n g . The b a r n a c l e can t h e n n e v e r a g a i n move. The g l u e w h i c h they use i s so s t r o n g t h a t , f o r example , i t i s e a s i e r to c h i p the r o c k t h a t the b a r n a c l e i s a t t a c h e d t o , than t o p u l l o f f the b a r n a c l e . I t has to be s t r o n g because of the f o r c e s t h a t the b a r n a c l e s have t o s t a n d up to i n the ocean s u r f . I f the b a r n a c l e c h o o s e s a bad p l a c e , where t h e r e i s not much f o o d , i t w i l l s t a r v e to d e a t h . I n any case i t might be e a t e n by p r e d a t o r s . The a d u l t b a r n a c l e l o o k s a l i t t l e l i k e a c lam c o v e r e d w i t h a p r o t e c t i v e s h e l l and w i t h a " s t a l k " pr "neck" w i t h which i t a t t a c h e s i t s e l f t o the o b j e c t . The s t a l k s e r v e s to keep the r e s t of the b a r n a c l e away f rom the r o c k . Organs which look a l i t t l e l i k e f e e t , p r o t r u d e f rom - 2 -the bottom of the s h e l l . It i s with these that the barnacle eats. The adult barnacle i s approximately 4-5 cm long. Mr. Harry Goldberg, for his master's thesis i n bioresource engineering at U.B.C. conducted an experiment r e l a t i n g to the s u r v i v a l and growth of barnacles. The barnacles were attached to l i n e s of three d i f f e r e n t types of material. These l i n e s were arranged i n two systems. Barnacles were dying due to lack of food and because of predators. At each of a number of times data was c o l l e c t e d on these barnacles, namely the number dead on each l i n e , s h e l l length, neck length, s h e l l weight and neck weight. Mr. Goldberg was interested i n modelling the s u r v i v a l d i s t r i b u t i o n and i n comparing the d i s t r i b u t i o n s for the three d i f f e r e n t types of l i n e and for the two systems. He was also interested i n f i t t i n g growth curves for each l i n e for s h e l l length and neck length and i n comparing the curves obtained across systems and material types. F i n a l l y he was interested i n the length-weight r e l a t i o n s h i p for s h e l l and neck. From our point of view the s t a t i s t i c a l methodology i s of more in t e r e s t than t h i s p a r t i c u l a r data set, so the length-weight r e l a t i o n s h i p , which involves straightforward regression, w i l l not be discussed. Also growth curves w i l l be f i t t e d only for s h e l l length, not for neck length. Everything that i s done for s h e l l length could be done i n an i d e n t i c a l way for neck length. Due to the l i m i t e d nature of the growth data - l o n g i t u d i n a l data was not c o l l e c t e d - i t was not straightforward to f i t meaningful growth curves and this presented i n t e r e s t i n g s t a t i s t i c a l problems. The remainder of Section 1 i s devoted to a f u l l e r d e s c r i p t i o n of the experiment and the data. In Section 2 the s u r v i v a l data i s - 3 -e x a m i n e d : f i r s t l y the s u r v i v a l d i s t r i b u t i o n s f o r v a r i o u s d i f f e r e n t l i n e s a r e compared u s i n g n o n p a r a m e t r i c t e s t s . Then the e x p o n e n t i a l model i s f i t t e d f o r each l i n e and the s l o p e s of the r e s u l t i n g c u r v e s a r e compared by means of the l i k e l i h o o d r a t i o t e s t . F i n a l l y random e f f e c t s mode ls a r e d e v e l o p e d f o r the s l o p e s of the c u r v e s . I n S e c t i o n 3 growth c u r v e s a r e f i t t e d f o r s h e l l - l e n g t h f o r each of the s e v e n t e e n l i n e s and v a r i o u s h y p o t h e s e s about the p a r a m e t e r s a re t e s t e d . In t h i s way i t i s d e t e r m i n e d whether sys tem and m a t e r i a l t ype a r e i m p o r t a n t f a c t o r s i n d e t e r m i n i n g growth c h a r a c t e r i s t i c s . However because of the l i m i t e d d a t a a v a i l a b l e , s t r o n g a s s u m p t i o n s have t o be made i n o r d e r to be a b l e t o a p p l y the growth cu rve m o d e l s . W i t h i n e a c h of the s e v e n t e e n l i n e s , t h e r e a r e a number of " i t e m s " ( w h i c h w i l l be d e s c r i b e d l a t e r ) . Growth c u r v e s a r e t h e n f i t t e d s e p a r a t e l y f o r each i t e m w i t h i n each l i n e . An a t tempt i s made to t e s t f o r d i f f e r e n c e s between the i t ems w i t h i n a l i n e , but t h i s i s found t o be i m p o s s i b l e w i t h the d a t a a v a i l a b l e . 1.1 The E x p e r i m e n t S e v e n t e e n l i n e s of t h r e e d i f f e r e n t t y p e s of m a t e r i a l were c o n s t r u c t e d as f o l l o w s : i . O y s t e r l i n e s - r o p e s , e a c h w i t h 10 or 11 o y s t e r s h e l l s a t t a c h e d . i i . D o w l i n g l i n e s - ropes each w i t h 20 p i e c e s of wood ( d o w l i n g ) a t t a c h e d . - 4 -i i i . N e t t i n g l i n e s - l o n g c y l i n d r i c a l p i e c e s of n e t t i n g each w i t h 10 compartments and a p i e c e of ha rd r u b b e r i n e a c h compartment . The s e v e n t e e n l i n e s were a r r a n g e d i n two s y s t e m s , the number of l i n e s i n e a c h sys tem b e i n g g i v e n be low: O y s t e r D o w l i n g N e t t i n g System 1 6 2 1 System 2 6 1 1 The l i n e s were t a k e n out t o the o c e a n , where b a r n a c l e s became a t t a c h e d to the i t e m s on each l i n e . The l i n e s were l a t e r r e t r i e v e d and se t up a l o n g the c o a s t l i n e i n a n a t u r a l e n v i r o n m e n t . At t h i s s t a g e no new b a r n a c l e s c o u l d become a t t a c h e d and t h o s e a l r e a d y a t t a c h e d c o u l d no t move. I n t h i s e n v i r o n m e n t the b a r n a c l e s were d y i n g due t o l a c k of f ood and because of p r e d a t o r s . A r e c o r d was made of the number of b a r n a c l e s i n i t i a l l y (a t t=0) on each l i n e . T h e r e were a p p r o x i m a t e l y 200 b a r n a c l e s i n i t i a l l y on each sys tem one o y s t e r l i n e , a p p r o x i m a t e l y 500 on each sys tem two o y s t e r l i n e , a p p r o x i m a t e l y 1000 on each d o w l i n g l i n e and on the system one n e t t i n g l i n e , and 1600 on the sys tem two n e t t i n g l i n e . The a c t u a l numbers a r e g i v e n i n T a b l e 1. At a number o f subsequent t imes d a t a was c o l l e c t e d on the b a r n a c l e s . - 5 -1.2 The Data S u r v i v a l Data A t t = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 14 and 17 weeks the number of b a r n a c l e s found dead on each l i n e was r e c o r d e d . The dead b a r n a c l e s were t h e n removed. So f o r example the b a r n a c l e s found dead at t = 7 weeks were t h o s e t h a t had d i e d between t = 6 and t = 7 weeks . No b a r n a c l e s d i e d b e f o r e t = 5 weeks . More d e t a i l e d i n f o r m a t i o n , namely the number of dea ths at each t ime on each i t e m w i t h i n a l i n e , was not a v a i l a b l e t o us a l t h o u g h i t was p r o b a b l y r e c o r d e d at some s t a g e d u r i n g the e x p e r i m e n t . The s u r v i v a l d a t a i s g i v e n i n T a b l e 1. Growth Data At each of the above t i m e s a h a p h a z a r d sample of l i v e b a r n a c l e s was s e l e c t e d f rom each i t e m on each l i n e . I t was assumed t h a t w i t h i n each i t e m each b a r n a c l e had e q u a l p r o b a b i l i t y of b e i n g s a m p l e d . F o r each sampled b a r n a c l e the f o l l o w i n g measurements were t a k e n : ( i ) neck l e n g t h ( i i ) s h e l l l e n g t h The d a t a was r e c o r d e d i n c e n t i m e t r e s . T h e s e b a r n a c l e s were not removed f rom the i t e m s . The sample s i z e was 5 e x c e p t i n the cases of samples f rom i t e m s on n e t t i n g l i n e s where i t was 10. So the t o t a l number of b a r n a c l e s measured at each t ime was: From o y s t e r l i n e s : 2 x 6 x 10 ( o r 11) x 5 = 600 From d o w l i n g l i n e s : 3 x 20 x 5 = 3 0 0 From n e t t i n g l i n e s : 2 x 10 x 10 = 200 So t h e r e i s an enormous amount of growth d a t a . The growth d a t a and the ways i n w h i c h i t i s i n c o m p l e t e , a r e d i s c u s s e d more f u l l y i n S e c t i o n 3 . - 6 -2. EXAMINATION OF THE SURVIVAL DATA In o r d e r t o o b t a i n the l e n g t h - w e i g h t d a t a , a p p r o x i m a t e l y 40 b a r n a c l e s were removed f rom each type o f m a t e r i a l a t each of a number of t i m e s (on a t o t a l of f i v e o c c a s i o n s ) . In a l l a n a l y s e s r e l a t i n g to the s u r v i v a l d a t a , t h i s f a c t was n e g l e c t e d . I t was assumed t h a t the remova l of t h e s e b a r n a c l e s would not s i g n i f i c a n t l y a f f e c t the a n a l y s e s . T h i s a s s u m p t i o n seemed r e a s o n a b l e s i n c e the b a r n a c l e s removed were chosen a t random and the number removed was v e r y s m a l l compared t o the t o t a l number: i n t o t a l a p p r o x i m a t e l y 200 b a r n a c l e s were removed f rom each type of m a t e r i a l , whereas the t o t a l number of b a r n a c l e s i n i t i a l l y on r e s p e c t i v e l y o y s t e r , d o w l i n g and n e t t i n g l i n e s was 6561, 2969 and 2552. The t o t a l number of b a r n a c l e s on each of t h e s e t y p e s of l i n e at the end o f the s tudy was r e s p e c t i v e l y 1668, 1200 and 1121. The i n i t i a l number of b a r n a c l e s on each l i n e and the number found dead a t each t ime a r e g i v e n i n T a b l e 1. No b a r n a c l e s d i e d b e f o r e t = 5 so the f i r s t t ime at w h i c h b a r n a c l e s were found dead was t = 6 . More d e t a i l e d i n f o r m a t i o n , namely the number of d e a t h s on each i t e m w i t h i n each l i n e was not a v a i a b l e to u s . The f i r s t q u e s t i o n of i n t e r e s t was whether the s u r v i v a l d i s t r i b u t i o n s f o r l i n e s of the same type w i t h i n the same sys tem were e s s e n t i a l l y the same. I f t h i s were found t o be the case i t would be r e a s o n a b l e to c o l l a p s e a c r o s s the s i x o y s t e r l i n e s w i t h i n each sys tem and a c r o s s the two d o w l i n g l i n e s i n sys tem 1. Then the s i t u a t i o n would be s i m p l e r - we would have t h r e e " t r e a t m e n t s " ( t y p e s of m a t e r i a l ) w i t h i n - 7 -Table 1 - Survival Data I n i t i a l Number Found Dead at Week # 6 7 8 9 10 11 14 17 01 204 31 32 33 3 6 4 7 5 02 187 27 30 22 0 8 3 11 9 03 134 22 31 15 1 3 18 1 3 04 137 27 21 19 5 6 7 5 5 05 187 34 26 29 9 6 10 4 8 06 306 40 27 25 9 7 14 11 14 DI 937 128 122 113 32 29 41 44 45 D2 817 127 168 118 37 32 16 50 37 Nl 952 75 226 119 116 19 75 10 7 07 530 30 42 69 34 35 26 39 20 08 180 9 18 12 7 14 7 9 17 09 579 24 40 39 47 78 49 37 17 010 155 14 16 28 9 9 1 7 4 011 677 20 61 80 35 62 30 67 16 012 470 7 22 33 28 38 21 33 25 D3 1215 50 89 126 77 96 61 99 32 N2 1600 19 23 118 216 175 110 75 48 - 8 -each of two systems instead of the present more complicated s i t u a t i o n of l i n e s within treatment within system. The l i n e e f f e c t would be known to be i n s i g n i f i c a n t . Three non-parametric rank tests were performed i n order to determine whether differences existed between the s u r v i v a l d i s t r i b u t i o n s of various groups of l i n e s . 2 .1 Description of the Non-Parametric Tests a. Logrank (Savage) Test This test can be derived either as a l i n e a r rank test or i n the context of Cox's proportional hazards model. It w i l l f i r s t be considered from the l a t t e r point of view: The proportional hazards model (Cox, 1972) for f a i l u r e time i s s p e c i f i e d by the hazard function X(t;z) = \o(t) exp (zp) where T i s f a i l u r e time, t i s the observed value of T, z ( l x p) i s a vector of covariates, B(p x 1) i s a vector of regression c o e f f i c i e n t s and Xo(t) i s an a r b i t r a r y unspecified base-line hazard function. Let U(p x 1) be the vector whose j t h element i s given by ologL j = 1 > PS where L i s the l i k e l i h o o d for 6. Let I(p x p) be the matrix whose ( j , h ) t h element i s given by j,h = 1 ,••• , p. - 9 -Then a t e s t of the n u l l h y p o t h e s i s H Q:8 = 8 0 i s based on the s c o r e s t a t i s t i c U(8o). Under Ho and o t h e r m i l d r e g u l a r i t y c o n d i t i o n s U(6o) i s a s y m p t o t i c a l l y n o r m a l w i t h mean 0 and c o v a r i a n c e m a t r i x w h i c h can be c o n s i s t e n t l y e s t i m a t e d by I(Bo). Note t h a t i s b e i n g used t o denote t h e o b s e r v e d F i s h e r i n f o r m a t i o n m a t r i x . To f i n d U and I the l i k e l i h o o d must be o b t a i n e d : L e t the d i s t i n c t d e a t h t imes be t ( ^ ) < . . . < t ( k ) * L e t dj[ be the number of dea ths a t t(±)* L e t n^ be the number a t r i s k j u s t p r i o r t o t ^ ^ . I n the case where t i e s and c e n s o r i n g a r e a l l o w e d the l i k e l i h o o d i s g i v e n by ( K a l b f l e i s c h and P r e n t i c e , 1980): k L = n i=i e x p C s ^ ) T. e x p C s ^ ) ^ R d i ( t ( i ) ) where : s^ i s the sum of the c o v a r i a t e s a s s o c i a t e d w i t h the d^ d i f a i l u r e s a t s^ = T, , 1 = (1^,... l d ) , ^ . ( t ^ ) i s the se t of a l l i n d i v i d u a l s a t r i s k at t ^ - 0 and ( t ( ^ ) ) i s t n e s e t of a l l i s u b s e t s of d^ i t e m s c h o s e n f rom R(t(£)) w i t h o u t r e p l a c e m e n t . The e x a c t p a r t i a l l i k e l i h o o d w i t h t i e s a r i s e s f rom a d i s c r e t e model s p e c i f i e d by X ( t ; z ) d t \ j ( t ) d t 1 - \ < t ; z ) d t = 1 - \ , < t ) d t e x p ( z P ) a - 1 0 -T h i s l i k e l i h o o d i s v e r y l a b o r i o u s to compute . P e t o ( 1 9 7 2 ) and B r e s l o w ( 1 9 7 4 ) s u g g e s t e d the f o l l o w i n g a p p r o x i m a t i o n to L: L = n e x p ( s , 8 ) / a i = l 1 R ( t ( i ) ) e x p ( z 8) T h i s a p p r o x i m a t i o n i s good p r o v i d e d t h a t d±/n± 1 S s m a l l , i = l , . . . , k . T h i s means t h a t a t e a c h t ime the number o f f a i l u r e s must be s m a l l compared to t h e number a t r i s k . o l o g L a > j ~ 1 , . . . , p. The t e s t s t a t i s t i c i s t h e n U ( 8 0 ) where U . ( B ) = 3 98j U(Bg) i s a s y m p t o t i c a l l y normal w i t h mean 0 and c o v a r i a n c e m a t r i x e s t i m a t e d by I ( 8 n ) where I . , (8) j n -b' l o g L 8 B j b K a ' j , h = 1 , . . . , p . In p a r t i c u l a r , to t e s t H o : 8 = 0 ( t h a t f a i l u r e t i m e s are u n r e l a t e d to the c o v a r i a t e s ) , U ( 0 ) ' V - 1 U ( 0 ) i s compared to X 2 ( p ) t a b l e s where U ( 0 ) = k i = l d i n i and V . , = I . , ( 0 ) , 3h j h I n the case where the s u r v i v a l d i s t r i b u t i o n s f o r samples f rom p + 1 d i f f e r e n t p o p u l a t i o n s a re to be compared, z ( l x p ) i s an i n d i c a t o r v a r i a b l e : z^ c o n s i s t s of a one i n the j t h p o s i t i o n and z e r o s e l s e w h e r e i f the i t h o b s e r v a t i o n i s i n the j t h s a m p l e . In t h i s case t e s t i n g H Q : 6 = 0 i s e q u i v a l e n t to t e s t i n g t h a t a l l the p o p u l a t i o n s have - l i -t h e same s u r v i v a l d i s t r i b u t i o n . In t h i s case k U . ( 0 ) = E ( d . . -J i = l J n j i d i / n i ) k d . ( n . - d . ) n, ' h i and = 1 ,••• ,p where 6 j n i s the K r o n e c k e r d e l t a ( 6 j n = 1 i f j = h and 0 o t h e r w i s e ) , n-ji = number a t r i s k i n the j t h sample j u s t p r i o r t o and d j ^ = number of d e a t h s a t t ime t ( i ) i t i t h e j t h s a m p l e . T h i s t e s t can a l s o be d e r i v e d as a l i n e a r rank t e s t f o r compar ing s u r v i v a l d i s t r i b u t i o n s w h i c h d i f f e r o n l y w i t h r e s p e c t t o l o c a t i o n ( L a w l e s s , 1982) : suppose we have p + 1 d i s t r i b u t i o n s d e f i n e d by p r o b a b i l i t y d e n s i t y f u n c t i o n s ( p . d . f . ' s ) g ( y - 9 i ) , . . . g (y - 0 p ) , g ( y ) where y = l o g l i f e t i m e . We w i s h t o t e s t t h a t 01 = . . . = 0 p = 0 i . e . t h a t a l l d i s t r i b u t i o n s a r e i d e n t i c a l . L e t z ( l x p) be an i n d i c a t o r v a r i a b l e ( Z - J J = 1 i f the i t h o b s e r v a t i o n i s f rom sample j and z^-j = Q o t h e r w i s e ) . L e t 0 = ( 0 1 , . . . , 0 p ) . T h e n , g i v e n the r e g r e s s i o n v e c t o r z , the p . d . f . of y i s f ( y | z ) = g ( y - z 0 ' ) . Under t h i s model we w i s h t o t e s t 0 = 0 . We f i r s t c o n s i d e r the case w i t h no c e n s o r i n g : suppose we o b t a i n a sample y i , . . . , y n f rom t h e s e d i s t r i b u t i o n s . L e t y ( i ) ^ ••• ^ y ( n ) ^ e the o r d e r e d o b s e r v a t i o n s (assumed to be d i s t i n c t ) and be the number of o b s e r v a t i o n s f rom d i s t r i b u t i o n i I f r (1 x n) i s the rank v e c t o r based on the y-j/s, then a t e s t of 0=0 can be based on U(0) (p x 1) whose j t h e lement i s g i v e n by ( i = l , . • •, P + D . - 12 -U.(0) = & l o y> ( r ; 0 ) where p ( r ; 0) i s the p r o b a b i l i t y mass f u n c t i o n of r . I t f o l l o w s t h a t U.(0) = £ Z ( i ) . a ± j - 1 , . . . ,p where z (^) i s the i n d i c a t o r v a r i a b l e a s s o c i a t e d w i t h y(^)> z ( i ) j i s the j t h e lement of z ( j j and the a-^'s a r e s c o r e s g i v e n by - g ' ( y ( i ) ) g ^ ( i ) ) To d e f i n e a p a r t i c u l a r rank t e s t , the o^'s a r e g e n e r a t e d by c h o o s i n g a s p e c i f i c p . d . f . g ( y ) . Then i f the d a t a a c t u a l l y comes f rom t h i s p . d . f . the t e s t w i l l be a s y m p t o t i c a l l y f u l l y e f f i c i e n t r e l a t i v e t o the p a r a m e t r i c p r o c e d u r e based on the a c t u a l v a l u e s r a t h e r t h a n the r a n k s of the y±'s. I f the d a t a a r i s e s f rom a d i f f e r e n t p . d . f . the rank t e s t w i l l be more e f f i c i e n t . In the case of the l o g r a n k t e s t the s c o r e s a r e g e n e r a t e d by l e t t i n g g ( y ) = e x p ( y - e v ) , the extreme v a l u e d i s t r i b u t i o n . The mean and c o v a r i a n c e m a t r i x f o r U(0) can be o b t a i n e d by p e r m u t a t i o n t h e o r y arguments ( L a w l e s s , 1 9 8 2 ) . The a^'s can be chosen so t h a t E ( U ( 0 ) ) = 0 . L e t V be the c o v a r i a n c e m a t r i x of U ( 0 ) . Then s i n c e U(0) i s a s y m p t o t i c a l l y n o r m a l , H O : 0 = O can be t e s t e d by compar ing U ( 0 ) ' V - 1 U(0) t o X 2 ( P ) t a b l e s . T h i s t e s t can be ex tended t o accommodate c e n s o r i n g as f o l l o w s : suppose t h a t t h e r e a r e k d i s t i n c t o b s e r v e d l o g l i f e t i m e s and n-k - 13 -c e n s o r i n g t i m e s . I f z(±) i s the i n d i c a t o r v a r i a b l e a s s o c i a t e d w i t h y ( i ) , l e t S ( i ) be the sum of t h e s e v e c t o r s f o r a l l i n d i v i d u a l s c e n s o r e d i n [ y ( i ) > v ( i + l ) ) « Then the s c o r e s t a t i s t i c s u g g e s t e d by P r e n t i c e (1978) i s U(0)(p x 1) where k V 0 ) ( z ( D j a i + s ( D j a i ) } J-I.--.P-So i n d v i d u a l s whose l i f e t i m e s a r e c e n s o r e d a r e g i v e n d i f f e r e n t s c o r e s a^ . The s c o r e s may be chosen so t h a t E[U(0)] = 0. I n p a r t i c u l a r , t a k i n g g ( y ) = e x p ( y - e^) as b e f o r e , and u s i n g P r e n t i c e ' s method of o b t a i n i n g the and a^ l e a d s t o the l o g r a n k t e s t v i a the s c o r e s 1 1 1 1 a i =.E,^7 - l> a i = .\ nT' where n^ = the number at r i s k j u s t p r i o r t o t ( ^ ) = e x p ( y ( . j j . ) Then we o b t a i n as b e f o r e k Uj(0) = - 2 ^ ( d j ± - n j ± d i / n 1 ) j = l , . . . , p , w i t h n o t a t i o n as b e f o r e . The e x p e c t a t i o n of U(0) i s 0. P r e n t i c e o b t a i n s a p e r m u t a t i o n v a r i a n c e f o r U(0). E i t h e r t h i s or the v a r i a n c e o b t a i n e d b e f o r e i n the c o n t e x t of C o x ' s model can be u s e d . S i n c e the s c o r e s were m o t i v a t e d by l e t t i n g g ( y ) = e x p ( y - e ^ ) , the extreme v a l u e d i s t r i b u t i o n , the t e s t i s a s y m p t o t i c a l l y f u l l y e f f i c i e n t f o r d e t e c t i n g l o c a t i o n d i f f e r e n c e s under an extreme v a l u e model f o r l o g l i f e t i m e s or e q u i v a l e n t l y , f o r t e s t i n g e q u a l i t y o f l i f e t i m e d i s t r i b u t i o n s i n a p r o p o r t i o n a l h a z a r d s or Lehmann f a m i l y when t h e r e i s no c e n s o r i n g o r e q u a l c e n s o r i n g i n a l l s a m p l e s . T h i s t e s t was d e r i v e d under the - 14 -a s s u m p t i o n of no t i e s but may be used w i t h a s m a l l number of t i e s . b. W i l c o x o n Test The W i l c o x o n ( o r as i t i s sometimes c a l l e d , P r e n t i c e ' s g e n e r a l i s e d W i l c o x o n t e s t ) can a l s o be d e r i v e d as a l i n e a r rank t e s t of the f o rm k U . ( 0 ) = ^ ( z ( i ) . a ± + s ( i ) . a ±) J - 1 . . . . . P f o r t e s t i n g e q u a l i t y of l i f e t i m e d i s t r i b u t i o n s . I n t h i s case the scores a r e d e f i n e d as ( P r e n t i c e , 1978) : a , = 1 - 2 n ( n . - d . + l ) / ( n . + 1) j =1 a , = 1 - n ( n . - d . + l ) / ( n . + 1) j = l J J J w i t h n o t a t i o n as b e f o r e . S u b s t i t u t i n g t h e s e s c o r e s i n U j ( u ) g i v e s n . . U . ( 0 ) = - E F , ( d . . - d , ) j = l , . . . , p j i j i i i ( n . - d . + 1) where F, = II — - ^ An e s t i m a t e of the c o v a r i a n c e m a t r i x of U(0) (see P r e n t i c e and Marek ( 1 9 7 9 ) ) i s V where k d . ( n . - d . ) n . . n, . TT y „2 i i i J i ( f , h i ^ V . , = £ F. — 7 - r r — - — 16., - - — I i , h = l , . . . , p . J h i (n^ - 1) n^ j h J J ' • , v - 15 -A g a i n U ( 0 ) ' V ^ l K O ) i s compared t o X 2 ( p ) t a b l e s . The s c o r e s i n t h i s case a r e g e n e r a t e d by t a k i n g g ( y ) = e^/( l + e y ) 2 , the l o g i s t i c d e n s i t y . So t h i s t e s t i s a s y m p t o t i c a l l y f u l l y e f f i c i e n t f o r d e t e c t i n g l o c a t i o n s h i f t s when the u n d e r l y i n g d i s t r i b u t i o n s a r e l o g i s t i c . Whereas the l o g rank t e s t g i v e s e q u a l w e i g h t t o a l l terms ( d j ^ - d ^ n ^ / n ^ ) , the W i l c o x o n g i v e s more w e i g h t to e a r l i e r e v e n t s than t o l a t e r o n e s . Hence t h i s t e s t i s good a t d e t e c t i n g d i f f e r e n c e s e a r l y o n . A g a i n i t was d e r i v e d under the a s s u m p t i o n of no t i e s , but may be used w i t h a s m a l l number of t i e s . c . L o g r a n k T e s t f o r Grouped D a t a I n the d e r i v a t i o n of the l o g rank t e s t i n the c o n t e x t of the p r o p o r t i o n a l h a z a r d s model i t was assumed t h a t i f t i e s were p r e s e n t t h i s was because the d a t a a r o s e f rom a d i s c r e t e m o d e l . I t may i n s t e a d be the c a s e t h a t the d a t a a r i s e s f rom a c o n t i n u o u s model but t h a t i t i s g rouped - the a c t u a l s u r v i v a l t ime i s not r e c o r d e d , o n l y the i n t e r v a l i n t o w h i c h i t f a l l s . T h i s would g i v e r i s e t o a s l i g h t l y d i f f e r e n t l i k e l i h o o d and thus t o a s l i g h t l y d i f f e r e n t t e s t than t h a t o b t a i n e d i n a , as w i l l now be d e s c r i b e d . C e n s o r i n g i s assumed o n l y to o c c u r j u s t p r i o r t o the end of an i n t e r v a l . Assuming a p r o p o r t i o n a l h a z a r d s model f o r the c o n t i n u o u s d a t a , t h e n , i f x± r e p r e s e n t s f a i l u r e i n t h e i t h i n t e r v a l , [ a ^ - i , a ^ ) , t h e h a z a r d c o n t r i b u t i o n a t x-£ f o r c o v a r i a t e z i s 1 - (1 - X, ) e X p ( z p ) , where (1 - X . ) = exp [ -/ 1 X (u) du] and X ( t ) 1 l a . . o o - 16 -i s the b a s e - l i n e h a z a r d f u n c t i o n . L e t y± = l o g [ - l o g ( l - \ ± ) ] , 1 = 1 , . . , k . Then the l i k e l i h o o d i s k L ( y , B ) = £ ( E l o g { l - e x p [ - e x p ( Y , + z B ) ] } - £ e x p ( y , + z B ) ) i = l 1€D ± 1 16 R 1 x where i s the se t of l a b e l s a t t a c h e d t o i n d i v i d u a l s f a i l i n g a t x^ and R^  i s the s e t of l a b e l s a t t a c h e d t o i n d i v i d u a l s c e n s o r e d a t x^ or o b s e r v e d t o s u r v i v e p a s t x^. A t e s t of Ho:8=0 i s based on u ( y ( 0 ) , 0) where U . ( y , 8 ) = -rs— l o g L and y(0) i s the maximum l i k e l i h o o d e s t i m a t e of y a t 6 = 0 . A g a i n u ( y ( 0 ) , 0) i s a s y m p t o t i c a l l y n o r m a l w i t h mean 0 and 2 c o v a r i a n c e m a t r i x e s t i m a t e d by l ( y ( 0 ) , 0) where ^JJJCYJP) = • Suppose t h a t t h e r e a re p + 1 s a m p l e s . When z i s an i n d i c a t o r v a r i a b l e f o r the samples i t f o l l o w s t h a t U J ( Y ( 0 ) , 0) = - E n±/d± l o g ( l - d . / n i ) ( d j ± - d ± n j ± / n ± ) j = l , . . . , p and the e l e m e n t s of the c o v a r i a n c e m a t r i x a r e q i V j h = E ( n 7 ) n j i ( 6 j h " n h i / n i ) j ' h = 1 P x i J n i ( n i " d i } r , „ d i ^ 2 where -l± = — a : ( l o g ( 1 - S 7 » and o t h e r n o t a t i o n i s as b e f o r e . The t e s t s t a t i s t i c i s U ( y ( 0 ) , 0 ) ' V - 1 U ( y ( 0 ) , 0) w h i c h can be compared t o X ( p ) 2 t a b l e s . n i d i I f d . / n . i s s m a l l t h e n - j — l o g ( l ) » -1 and the t e s t i s i l i i - 17 -a p p r o x i m a t e l y the same as the l o g r a n k t e s t i n a . A p p l i c a t i o n o f t h e T e s t s I n our case we have c o n t i n u o u s d a t a wh ich has been grouped - the a c t u a l s u r v i v a l t i m e s a r e not r e c o r d e d , o n l y the i n t e r v a l i n t o w h i c h they f a l l . T h i s g i v e s r i s e to a l a r g e number of t i e s . However t h e r e a r e a l s o a l a r g e number at r i s k and d^/n^ i s r a r e l y b i g g e r than 0.2 and u s u a l l y much s m a l l e r . C e n s o r i n g i s due o n l y to the f a c t t h a t some b a r n a c l e s a r e s t i l l a l i v e a t the end o f the s t u d y - t h i s i s t ype I c e n s o r i n g . So we have the same c e n s o r i n g p a t t e r n i n a l l s a m p l e s . The u n d e r l y i n g d i s t r i b u t i o n of the s u r v i v a l t i m e s i s not known so i t i s not c l e a r whether the l o g r a n k t e s t o r the W i l c o x o n t e s t w i l l be more p o w e r f u l . We do know, however , t h a t the W i l c o x o n i s more s e n s i t i v e t o d i f f e r e n c e s e a r l y on whereas the l o g r a n k i s more s e n s i t i v e t o d i f f e r e n c e s l a t e r . I f t h e l o g r a n k t e s t i s t o be used i t would be more a p p r o p r i a t e t o use the t e s t d e s c r i b e d i n c . as we have grouped d a t a . The l o g r a n k t e s t a . on t h e o t h e r hand assumes a d i s c r e t e m o d e l . A l s o the a p p r o x i m a t i o n to the l i k e l i h o o d used i n the d e r i v a t i o n of t h i s t e s t assumed s m a l l v a l u e s of d-^/n^. I t i s not c l e a r whether our v a l u e s o f d^/n^ a r e s u f f i c i e n t l y s m a l l . The W i l c o x o n was d e r i v e d under the a s s u m p t i o n of no t i e s and i s o n l y an a p p r o x i m a t i o n when, as i n our c a s e , a l a r g e number o f t i e s a r e p r e s e n t . \ - 18 -E a c h t e s t was a p p l i e d t o v a r i o u s g rous of l i n e s t o t e s t the e q u a l i t y of the s u r v i v a l d i s t r i b u t i o n s of the l i n e s , w i t h the f o l l o w i n g r e s u l t s : L o g r a n k f o r L i n e s Compared L o g r a n k W i l c o x o n grouped d a t a 01,02,03,04,05,05 2 X = 5 34 .7 2 X = 5 31 .8 2 X = 5 3 5 . 9 07,08,09.0io,0n,0i2 2 X = 5 2 8 . 9 2 X = 5 36 .6 2 X = 5 3 3 . 6 D l , D 2 2 X = 1 32 .5 2 X = 1 29 .3 2 X = 1 32 .9 Dl.D2.D3 2 X = 2 129.8 2 X = 2 164.2 2 X = 2 126.7 Ni,N 2 2 X = 1 129.8 2 X = 1 301.6 2 X = 1 196.2 D 3 , N 2 2 X = 1 5.8 2 X = 1 12.6 2 X = 1 5.1 A l l p - v a l u e s a r e < .001 e x c e p t t h a t o b t a i n e d i n the c o m p a r i s o n of D3 and N2. Here the l o g r a n k t e s t f o r grouped d a t a g i v e s p = . 0 2 5 , the l o g r a n k g i v e s p - .017 and the W i l c o x o n g i v e s p < . 0 0 1 . F o r each group of l i n e s , the h y p o t h e s i s t h a t t h e i r s u r v i v a l d i s t r i b u t i o n s a r e the same i s s t r o n g l y r e j e c t e d . The p v a l u e s a r e v e r y s m a l l p o s s i b l y due t o the l a r g e amount of d a t a . The s m a l l e s t p - v a l u e s o c c u r when l i n e s f rom d i f f e r e n t s y s t e m s , f o r example Ni and N2, a r e compared . On the o t h e r hand the s u r v i v a l d i s t r i b u t i o n s f o r D2 and N3, l i n e s of d i f f e r e n t t y p e s w i t h i n the same s y s t e m , appear t o be much more s i m i l a r j u d g i n g by the - 19 -r e l a t i v e l y s m a l l c h i - s q u a r e d v a l u e . T h i s s u g g e s t s t h a t sys tem may be a more i m p o r t a n t f a c t o r than m a t e r i a l t ype i n d e t e r m i n i n g the s u r v i v a l d i s t r i b u t i o n . In any case d i f f e r e n c e s e x i s t even between l i n e s of the same type w i t h i n a sys tem and i t i s c l e a r l y not r e a s o n a b l e to c o l l a p s e o v e r t h e s e g r o u p s . The d i s c r e p a n c y between the r e s u l t s of the two l o g r a n k t e s t s i s s u r p r i s i n g l y s m a l l e x c e p t i n the c o m p a r i s o n of N i and N2. T h i s c o u l d be b e c a u s e d-j/n^ i s a lways q u i t e s m a l l i n s p i t e o f the l a r g e number of t i e s . F o r the f i r s t t h r e e compar i sons the W i l c o x o n g i v e s a s i m i l a r r e s u l t to the l o g r a n k t e s t s but f o r the l a s t t h r e e compar i sons i t g i v e s a much s m a l l e r p - v a l u e than the o t h e r two t e s t s . T h i s c o u l d r e f l e c t the f a c t t h a t the W i l c o x o n i s b e t t e r at d e t e c t i n g d i f f e r e n c e s e a r l y o n , w h i c h i s where the b i g g e s t d i f f e r e n c e s l i e i n t h e s e c a s e s . F o r example the p e r c e n t a g e d y i n g between weeks 6 and 7 on l i n e N i i s 23.7 whereas on N 2 t h i s p e r c e n t a g e i s 1 .4 . L a t e r on the d i s c r e p a n c i e s a r e not as g r e a t . 2.2 E x p o n e n t i a l M o d e l s H a v i n g compared the s u r v i v a l d i s t r i b u t i o n s f o r d i f f e r e n t l i n e s n o n p a r a m e t r i c a l l y , i t was t h e n o f i n t e r e s t t o l o o k f o r a p a r a m e t r i c model t o f i t the s u r v i v a l c u r v e s . T h e r e were two r e a s o n s f o r d o i n g t h i s . F i r s t l y the shape of the s u r v i v a l c u r v e s was of i n t e r e s t i n i t s e l f and s e c o n d l y under a p a r a m e t r i c model the t a s k of compar ing s u r v i v a l d i s t r i b u t i o n s f o r d i f f e r e n t l i n e s would be s i m p l e r - i n s t e a d of h a v i n g to compare a l a r g e amount of d a t a , namely number of d e a t h s and - 20 -number at r i s k at each time, we would j u s t have to compare a small number of parameters, for example slope. With so much data i t was l i k e l y that any parametric model would be rejected by a goodness-of-fit t e s t . This shouldn't matter provided that the model captures the most important features of the data. Because of i t s s i m p l i c i t y the exponential was the f i r s t choice of model. The exponential was f i r s t f i t t e d to each of the 17 l i n e s s t a r t i n g at t=0. So the following model was assumed: P(T > t) = -Bt e t > 0 0 otherwise where T = l i f e t i m e of barnacles. For each of the 17 l i n e s the maximum l i k e l i h o o d estimate of 8 was obtained. Notation i s defined as follows: to=0 ti=5 t2=6 ts=14 tg=17 d^ deaths d 2 deaths d 9 deaths N-Zd^ Under t h i s model the l i k e l i h o o d i s proportional to L - (1 - e ^ f l ( e ^ l - e - e ^ .... - e'**')*9 ( e " ^ 9 Since there were no deaths p r i o r to t i = 5 weeks on any of the l i n e s , d i = 0 and the f i r s t term i s i d e n t i c a l l y one. The l a s t term i s the contribution of the censored observations: under our model the p r o b a b i l i t y that a barnacle survives at least u n t i l the end of the study - 21 -at t 9 = 17 i s e ' P ^ . The number that do survive t h i s long i s N-Edi where N i s the i n i t i a l number on the l i n e . To obtain the maximum l i k e l i h o o d estimate of B, note that -Bt L -Bt I d i ^ i - l e + t i e ) = Z ~ * ' - B t TBT t 9 ( N - Ed.) 96 i = 2 Bt j Bt x (e - e ) a 2 i o g L I . , -P<V*i-i> ^ i - i - , E d |-e }. (e - e ) 2 ^ ^ o g L i s c l e a r l y < 0 for a l l 8 since the d are >^  0 (and c l e a r l y at oB 2 l e a s t one d i s > 0). So the turning point at the so l u t i o n to ^ ^ ° g L = 0 1 Op i s 8, the maximum l i k e l i h o o d estimate of p. This was obtained by doing a Newton Raphson i t e r a t i o n : P J .1 - P - U p )/ UB ) Mn+1 r n r n K n and 8 L = 1/T where T = median s u r v i v a l time (not exact due to discreteness) and ' _ ologL _ o 2logL When convergence to desired accuracy i s achieved -L(B) i s the observed Fisher information and leads to an estimate of the variance of B: Var 8 = [ - U p ) ] - 1 , SE(B) = [Var p ] 1 / 2 A f t e r obtaining 6, two goodness of f i t tests were ca r r i e d out to check - 22 -the f i t of the model for each line. 1. Pearson's Goodness of F i t Test k+1 (0 - E ) 2 X = £ — i = i i=l E i where 0^  = observed number of deaths in i t h interval and E^ = expected number of deaths in i t h interval -8t -Bt Under our model the estimate of is E^ = N(e - e ) and k + 1 2 2 is the number of time intervals. X i s compared to X (k- s) tables where s = number of parameters estimated in the model (=1). If X is too large the f i t is poor. It wasn't necessary to combine any intervals as for a l l lines the expected frequency in a l l intervals was at least 5. 2. L i k e l i h o o d Ratio Goodness of F i t Test k+1 0, 2 i 2 X = 2 S 0.1og(—) is compared to X,. . tables where i=l 1 h± Ck-s; notation i s as for 1. Results Results are given in Table 2. Comparing to chi-squared tables with eight degrees of freedom, the p-values for the goodness-of-fit tests are a l l < .001. Clearly the model doesn't f i t . The largest residuals were found to be at the beginning which is not surprising since the model gives a positive probability of death for t€ [0,5] but no deaths were T a b l e 2 - R e s u l t s of F i t t i n g E x p o n e n t i a l Mode l S t a r t i n g a t t = 0 . 8 ± SE(8) P e a r s o n ' s X 2 L i k e l i h o o d R a t i o X 2 0 i 0 .0520 + 0.0047 299.3 242 .7 0 2 0.0506 + 0.0048 221 .5 192.4 0 3 0.0672 + 0.0069 263.7 216 .8 0 .6540 + 0.0067 187.7 167 .0 05 0.0623 + 0.0055 251 .9 226 .2 o6 0.0374 + 0.0031 232.2 2 1 5 . 5 D l 0 .0500 + 0.0021 368.1 380 .4 D 2 0.0680 + 0.0028 94 .6 106.8 N i 0 .0638 + 0.0025 506.4 4 7 7 . 0 0 7 0.0444 + 0.0026 183.3 154.6 o8 0.0391 + 0.0041 500.1 508 .6 0 9 0.0455 + 0.0025 218.7 238 .8 OlO 0 .0475 + 0.0051 905.4 8 4 4 . 5 O i l 0 .0432 + 0.0022 1168.2 1008.0 0 l 2 0 .0318 + 0.0022 743 .9 402 .8 D3 0 .0402 + 0.0016 1692.0 1419 .0 N 2 0.0367 + 0.0013 1707.3 1976.0 - 24 -a c t u a l l y observed i n t h i s period. So the same model was then f i t t e d s t a r t i n g at t=5: P(T > t) --B(t-5) t > 5 « otherwise L e t t i n g t' = t - 5, we have the following s i t u a t i o n t ' - 0 t{ - 1 t' - 9 t' - 12 di deaths d 2 deaths d 8 deaths N-Zd For t h i s model the l i k e l i h o o d i s -Bt' d -BtJ -Bt' d L = (l - e ) (e - e ' ,"2 , "B t J - B t8> d8 , - B t 8 ^ M l J [e " e J (e J where t j =1, t£ = 2 t£ = 6, t j =9, t£ = 12. Now i v i - i  -at-1 i=2 so T 8 d.C-t! , e ologL _ v i - i 96 " 4_. . -Bt! -Bt;., -pt' -BtJ + t' e (e - e J Z | t T - tJCN - Ed.). (1 - e ') The new maximum l i k e l i h o o d estimates of p were obtained and the goodness-of-fit tests repeated. Results appear i n Table 3. The p-values f o r the tests are much larger t h i s time but are again a l l < .001. The f i t i s much better but s t i l l poor. The model appears to f i t better for the oyster l i n e s than for the dowling or netting l i n e s , though t h i s might be because there i s a smaller amount of data for the oyster l i n e s . It was possible that a better f i t might be obtained by s t a r t i n g at some point other than t=5. So the same model was f i t t e d again, t h i s time T a b l e 3 - R e s u l t s of F i t t i n g t h e E x p o n e n t i a l M o d e l S t a r t i n g a t t = 5 B ± SE(B) 2 P e a r s o n ' s X L i k e l i h o o d R a t i o X2 O i 0 .0927 + 0.0084 94 .5 103.0 0 2 0.0883 + 0.0084 6 2 . 5 6 8 . 9 03 0.1290 + 0.0130 77 .2 9 4 . 3 0 4 0.1240 + 0.0130 40 .6 4 5 . 9 05 0 .1160 + 0.0100 59 .3 69 .7 o6 0.0613 + 0.0050 69 .2 7 0 . 5 0 .0867 + 0.0037 80 .8 7 9 . 8 D 2 0.1298 + 0.0054 17.9 18.3 N l 0 .1200 + 0.0047 148.1 141.9 0 7 0.0740 + 0.0043 58 .3 59 .2 o8 0.0630 + 0.0065 127.7 136.9 0 9 0.0755 + 0.0042 54 .1 53 .6 OlO 0 .0818 + 0.0087 221 .8 235 .0 O i l 0 .0713 + 0.0037 242.5 253 .5 0l2 0.0497 + 0.0034 171.8 187.8 D 3 0.0656 + 0.0026 495 .5 5 9 9 . 4 N 2 0.0588 + 0.0021 681 .4 6 3 9 . 8 - 26 -with two unknown parameters - the slope parameter 6 and the loc a t i o n parameter a - both of which are to be determined by maximum l i k e l i h o o d . Since the l i k e l i h o o d i s d i f f e r e n t for a i n d i f f e r e n t time i n t e r v a l s , to A A f i n d a and 8 i t i s necessary to determine a - p r i o r i i n which i n t e r v a l the maximum l i k e l i h o o d estimate of a l i e s . A Fortunately i t was possible to determine a - p r i o r i that a would have to l i e i n the i n t e r v a l (5,6] assuming no deaths i n (0,5] and at least one death i n (5,6). In our case these conditions are s a t i s f i e d since no deaths are observed i n (0,5] but on every l i n e the f i r s t deaths are observed i n (5,6]. A Proof that a € (5,6] For a e (0,5] and no deaths i n (0,5], the l i k e l i h o o d i s i-1 8 -8 ( t 9 - a ) ( N - E d.) i=l 1 e 0 t i - 5 t2=6 t 8-14 t 9=17 0 deaths d^ deaths d 8 deaths N-Ed. Then 8 9 logLj = BctN - 8t 9(N - E d ) + E d . log(e i= l 1 i=2 1 - 1 -Bt i-1 - e ) = G( 8) + pctN - 27 -where G(B) i s a f u n c t i o n of 6 i n d e p e n d e n t of a . S i n c e 8 > 0 , l o g L i i s monoton i c i n c r e a s i n g i n a f o r a £ ( 0 , 5 ] . So a >_ 5. F u r t h e r m o r e a <^  6 s i n c e a > 6 would mean t h a t t h e r e would be a z e r o p r o b a b i l i t y of d e a t h i n ( 5 , 6 ] , but we have assumed t h a t deaths a r e o b s e r v e d i n t h i s i n t e r v a l , a ^ 5 and a _< 6 t o g e t h e r i m p l y a € ( 5 , 6 ] . m. Determination of a F o r a £ ( 5 , 6 ] the l i k e l i h o o d i s 8 - 8 ( t 2 - a ) J 9 - 6 ( t , , - o ) - B ( t . - a ) d . . - B ( t 9 - a ) ( N - £ d , ) L 2 = (1 -e ) d l n (e 1 1 - e 1 ) 1 1 e 1 1 i=3 to-o t x = 5 t 2 = 6 t 8 = 1 4 t 9 = 1 7 I 4 i i i di d e a t h s dg dea ths N-Ed i S i n c e we have a l r e a d y a s c e r t a i n e d t h a t oc€ ( 5 , 6 ] , t h i s i s the a p p r o p r i a t e l i k e l i h o o d . ^i£|i2 = B (N - dV<l - e - P ( T 2 _ A ) ) ) ?,2, T Q 2 , - p ( t 2 ~ a ) and 9 l o g L 2 = "P d l e . 2 da ^ _ e - 8 ( t 2 - a ) j 2 2 S i n c e & l o g L 2 < 0 f o r a l l a and 8 > 0, the s o l u t i o n t o a i ° g L 2 = 0 a 2 ba da - 28 -y i e l d s a l o c a l maximum. = 0 can be s o l v e d e x p l i c i t l y f o r a i n terms o f 8: So i f we maximise L 2 over a and 8 s i m u l t a n e o u s l y we o b t a i n 1 ,_rN - d P where 8 i s the maximum l i k e l i h o o d e s t i m a t e of 6. S i n c e t2=6 , 8 > 0 and l o g f — < 0, we have a < 6 . Case 1 I f a e ( 5 , 6 ] , (<===> - l n ( N " d * ) 6 ( - 1 , 0 ] ) t h e n a i s the m * N m 8 maximum l i k e l i h o o d e s t i m a t e , a , o f a , s i n c e i t i s the s o l u t i o n t o k 2 . SlogL ^  = 0 , 0 l o g L ? ^ Q a n ( j L 2 i s the a p p r o p r i a t e l i k e l i h o o d f o r a da ' - 2 oa i n t h i s i n t e r v a l . Case 2 I f a < 5 (<===> — l n ( N „ d J-) < - 1 1 , t h e n a i s not the maximum m v * N ' m P 9a not the a p p r o p r i a t e l i k e l i h o o d f or a K 5. I n t h i s case the t r u e maximum l i k e l i h o o d e s t i m a t e of a . I t i s the s o l u t i o n to & 1 ° f L 2 = 0 but L 2 i s l i k e l i h o o d e s t i m a t e i s a = 5 . - 29 -Proof For given p, ^ ^ ^ 2 . = Q n a s a u n i q u e sol u t i o n , a m , which i s assumed to l i e i n (0,5]. We have shown that the turning point i s a maximum and therefore logL>2 i s monotonic decreasing i n a to the right of a m and i n A A /\ p a r t i c u l a r for a € (5,6], So a <^5. But we know that a & (5,6], A So a = 5. Determination of 8 logL2 was i n i t i a l l y maximised over a and 8 simultaneously to obtain and 8 m. This was done v i a a Newton Raphson i t e r a t i o n as follows: Let L = 5 1 ° g L 2 L = a i °S L2. U e Z a 5a ' 8 58 » = Q 2logl2 L = a 2 l ° g L 2 L = a 2 l o g L 2 . Q a 2 » ap 5a5p ' pp b?2 I n i t i a l estimates were given for a and p. Then the i t e r a t i o n -1 (a ,p ) = (a ,8 ) - (L ,Lfl) - -n+1 Kn+1 n' rn a p L * ( a n ' P n ) L L aa ap L a p Lpp /\ A (VBn> was used u n t i l convergence to desired accuracy was achieved. Case 1 Suppose we obtain a m £ (5,6]. Then a and p are the true maximum l i k e l i h o o d estimates a and p. m m In t h i s case an estimate of the covariance matrix of (a, p) i s given by: - 30 -L L o" aa aB L a B L B B (a ,P) _1 t h e i n v e r s e of t h e o b s e r v e d F i s h e r i n f o r m a t i o n . I n our c a s e , s i n c e A 2 = 0 can be s o l v e d e x p l i c i t l y t o f i n d a = a ( B ) , we c o u l d i n s t e a d have s u b s t i t u t e d f o r a i n terms of 8 i n ^°f^^ = 0 . Then we would have op had j u s t one e q u a t i o n i n one unknown. T h i s c o u l d have been s o l v e d by a A Newton Raphson i t e r a t i o n to f i n d 8 and i t s s t a n d a r d e r r o r . Then A A A a = a(B) and the v a r i a n c e of a would have been e s t i m a t e d by ~ L 8 8 ( L a a L B 8 " L Case 2 ' 0 2 ) _ 1 | * * p (a,6) Suppose we o b t a i n a_ < 5. Then a and 8 a r e not the t r u e r m m "m maximum l i k e l i h o o d e s t i m a t e s . As a l r e a d y shown, a = 5. Knowing t h a t A a = 5, we can t h i n k o f a as f i x e d . We max imise w i t h r e s p e c t to 8 g i v e n t h a t a = 5. T h i s means t h a t when we f i t t e d the e x p o n e n t i a l s t a r t i n g a t t = 5, we a l r e a d y had the b e s t p o s s i b l e f i t as t = 5 was the b e s t p o i n t A a t w h i c h to s t a r t . So 8 i s as o b t a i n e d i n t h i s p r e v i o u s a n a l y s i s . As 2N-1 f o r case 1 t h e v a r i a n c e of 6 i s e s t i m a t e d by - L (L L„„ - L „ )" K aa aa pp ap (a ,B) The X v a l u e s a r e the same as t h o s e o b t a i n e d i n the p r e v i o u s a n a l y s i s s t a r t i n g a t t = 5. We have one degree of f reedom fewer but t h i s s h o u l d n ' t a f f e c t the c o n c l u s i o n s . - 31 -R e s u l t s A In t h o s e c a s e s f o r w h i c h a > 5 t h e f i t was s l i g h t l y improved by m a x i m i s i n g over two p a r a m e t e r s . T h i s improvement was r e f l e c t e d i n t h e s l i g h t l y s m a l l e r X v a l u e s . F o r seven of the s e v e n t e e n l i n e s i t was found t h a t t = 5 was not the i d e a l p l a c e to s t a r t . The r e s u l t s f o r t h e s e l i n e s a r e as f o l l o w s : A A A A L i k e l i h o o d R a t i o 6 ± S E ( 8 ) a ± S E ( a ) P e a r s o n ' s X 2 X 2 o 7 0 .0763 + .0047 5.24 ± .147 76 .1 7 7 . 8 o 8 0.0646 + .0071 5.21 ± .279 16 .9 17 .8 0 9 0 .0805 + .0046 5.47 ± .112 129 .2 130 .8 O i l 0 .0774 + .0041 5.61 ± .089 99 .7 114 .2 O 1 2 0 .0541 + . 0038 5.72 ± .107 3 7 . 3 36 .7 N i 0.1281 + .0054 5 .36 ± .079 454 . 2 584 .8 N 2 0 .0624 + .0019 5.81 ± .152 682 .1 580 .3 V a l u e s of a a r e g e n e r a l l y s m a l l e r f o r the system one o y s t e r l i n e s and f o r the d o w l i n g l i n e s than f o r o t h e r l i n e s . I t a p p e a r s t h a t b a r n a c l e s on t h e s e l i n e s s t a r t d y i n g s o o n e r . However the s c a l e pa ramete r 8, w h i c h r e l a t e s t o the r a t e a t w h i c h the b a r n a c l e s d i e o f f , i s of more i n t e r e s t . 2.3 Inference on the p 's A A l t h o u g h the model d o e s n ' t f i t v e r y w e l l , 8 i s a good summary s t a t i s t i c and c o n t a i n s a l o t of i n f o r m a t i o n about the l i n e s . So many - 32 -a n a l y s e s r e l a t i n g to the 8 v a l u e s were then c a r r i e d o u t . A l l of t h e s e a n a l y s e s were based on the e s t i m a t e s of 6 o b t a i n e d by m a x i m i s i n g o v e r two p a r a m e t e r s . P r e v i o u s l y we had compared the s u r v i v a l d i s t r i b u t i o n s o f c e r t a i n g roups o f l i n e s n o n p a r a m e t r i c a l l y . Now we were i n a p o s i t i o n to compare the d i s t r i b u t i o n s of the same groups of l i n e s p a r a m e t r i c a l l y by t e s t i n g whether a l l s y s t e m one o y s t e r l i n e s ( s a y ) had the same u n d e r l y i n g 8 v a l u e . T h i s was done by means of the l i k e l i h o o d r a t i o t e s t as f o l l o w s : To t e s t H 0: 81=82=83=64=65= 86 o 2[logL(j3, a) - logL(j3 0, SLO^ i s compared t o X t a b l e s and H Q i s r e j e c t e d i f t h i s s t a t i s t i c i s t o o l a r g e . Here B j d e n o t e s t h e t r u e s l o p e f o r t h e * j t h o y s t e r l i n e on sys tem one . T h i s t e s t i s v a l i d p r o v i d e d t h a t samples a r e l a r g e , w h i c h they a r e . Here q = number o f A A c o n s t r a i n t s imposed by Ho , l o g L ( j 3 , a) = v a l u e of l o g l i k e l i h o o d a t t h e A A maximum l i k e l i h o o d e s t i m a t e s and logL(j3o> jxo) = v a l u e of l o g l i k e l i h o o d a t the maximum l i k e l i h o o d e s t i m a t e s o b t a i n e d under HQ. In our case L(j? , a) i s the l i k e l i h o o d of o b t a i n i n g the o b s e r v e d numbers of dea ths on e a c h l i n e i f the j t h l i n e f o l l o w s the e x p o n e n t i a l model w i t h p a r a m e t e r s Bj and otj w h i c h may be d i f f e r e n t f o r each l i n e . A A So t o o b t a i n logL(j3, jx) we maximize f o r each l i n e s e p a r a t e l y ( w h i c h we have a l r e a d y done) and add the maximum v a l u e s of the l o g l i k e l i h o o d . L(_B_o» J*o) I s the l i k e l i h o o d of o b t a i n i n g the o b s e r v e d numbers o f d e a t h s on e a c h l i n e i f the j t h l i n e f o l l o w s the e x p o n e n t i a l model w i t h p a r a m e t e r s 8 and a * . So a l l l i n e s must have a common B but t h e i r - 33 -l o c a t i o n p a r a m e t e r s ai,...,06 may be d i f f e r e n t . T h i s t ime the s i x p a r t s cannot be maximized s e p a r a t e l y . 9 ~ B t 1 _ 1 - B t . 8 l o g L ( 8 0 , <xo) = £ d , , l o g ( e - e ) - 6 t9(N - E d . ) i=3 1 1 i = l 1 + B [ a i (N i - d u ) + a 2 ( N 2 - d i 2 ) + . . . + a 6 ( N 6 - d i e ) ] + d n l o g ( l - e - e ( t 2 _ a i ) ) + . . . + d 1 6 l o g ( l - e " B ( t 2 " a 6 ) ) where Nj = i n i t i a l number of b a r n a c l e s on l i n e j , d^j = number of d e a t h s a t t ime t^ on l i n e j , 6 d^= £ d = t o t a l number of dea ths at t ime i and N=Nj^  + N 2 + . . + N 6 . j = l In t h i s c a s e , = 0 can be s o l v e d t o g i v e i , N . - d, . a. = t 2 + j l o g ( J N J ) j = l , 2 , . . . , 6 . 2 T h i s i s c l e a r l y a maximum s i n c e — < 0 f o r a l l a , and 8 > 0 . T h i s can be s u b s t i t u t e d i n 9 ^ ° g L t o g i v e op - B t - B t , a i L 8 9 d . , ( t , e 1 - t e 1 A ) - ^ f - = - t 9 ( N - E d ) + E i~- + t 2 ( N - d l ) . °P 1 = i 1 i = 3 8 t i _ l P t 1 (e - e ) A Newton Raphson i t e r a t i o n u s i n g < ^ S k = I d l o B 2 i=3 1 - e h-i - c i - 6 t i - i - B t i e - e - 34 -w h i c h i s c l e a r l y < 0 then y i e l d s 8. I t i s t h e n s t r a i g h t f o r w a r d t o A o b t a i n a j , j = l , 2 , . . . , 6 and t o s u b s t i t u t e the maximum l i k e l i h o o d A A e s t i m a t e s i n t o logL(J3o >ato) to o b t a i n logL(8_o, ap). R e s u l t s 1. T e s t H Q : t h a t a l l sys tem one o y s t e r l i n e s have the same 8 v a l u e . A A A A 2[ logL(£ , a) - l o g L ( £ 0 , o 0 ) ] = 2( -2190 - ( - 2 2 5 7 ) ) = 135. R e f e r t o X 2 5 t a b l e s . P « . 0 0 1 . 2. T e s t Ho: t h a t a l l sys tem two o y s t e r l i n e s have the same 8 v a l u e . A A A A 2 [ l o g L ( p , o) - l o g L ( j 5 0 , «0)1 = 2 ( - 4 7 4 0 - ( - 4 7 5 3 ) ) = 2 6 . R e f e r t o X 2 5 t a b l e s . P « . 0 0 1 . 3. T e s t H Q : p D ^ = p ^ = Pn 3> t h a t a l l the d o w l i n g l i n e s have the same P v a l u e . A A A A 2 [ l o g L ( j 3 , a) - l o g L ( p o , oo)] = 2( -5648 - ( - 5 7 5 9 ) ) = 222. R e f e r t o X 2 2 t a b l e s . P « . 0 0 1 . 4 . T e s t Ho: p n = p n , t h a t the sys tem one d o w l i n g l i n e s have t h e u l u2 same P v a l u e . S i n c e we a r e compar ing o n l y two p v a l u e s , a t - t e s t can be u s e d , p r o v i d e d t h a t the sample f rom w h i c h p,, and p_ a r e e s t i m a t e d i s u l u2 r e a s o n a b l y l a r g e . I n our case i t i s v e r y l a r g e . The use of a t - t e s t t can be j u s t i f i e d as f o l l o w s : i n g e n e r a l , i f 9 i s a maximum l i k e l i h o o d A e s t i m a t e o f 9, t h e n 9 i s a s y m p t o t i c a l l y norma l w i t h mean 9 and - 35 -9 2 2 covariance matrix {E[ ] } - 1 which can be estimated by {—5—i£li. 0 90 2 50 2 This arises from the central l i m i t theorem applied to the sum of the A A contributions to the log l i k e l i h o o d . Since £L. and 6_ are maximum D l D2 l i k e l i h o o d estimates and since t h e i r variances are estimated, under HQ 8 " " 8 and for large samples, ^1 ^2 might reasonably be A A / ( S E ( 8 D i ) 2 + s*%y) compared to t-tables. This s t a t i s t i c works out to 6.6 and HQ i s rejected with p < .001. 5. Test Hn: 8XT = 8„ , that the two netting l i n e s have the same 8 N i N 2 value. = 11.5 i s referred to a t - d i s t r i b u t i o n . Again, A A 2 / ( S E ( 8 N i ) + S E ( 8 N 2 ) ^ p < .001 and HQ i s rejected. Differences were found i n every set of 8 values. The 6 values for the system two oyster l i n e s have a much smaller range of values than the 2 values for the system one oyster l i n e s as indicated by the smaller X value. The non-parametric tests also found the s u r v i v a l d i s t r i b u t i o n s to be d i f f e r e n t for d i f f e r e n t l i n e s but found the system one oyster l i n e s and the system two oyster l i n e s to be approximately equally homogenous i n t h e i r s u r v i v a l d i s t r i b u t i o n s . This discrepancy could a r i s e from the fact that our parametric model doesn't f i t the data very w e l l . - 36 -Other Parametric Models The Weibull would have been the natural choice of a model to f i t next. Since t h i s model has two parameters (or three i f a l o c a t i o n parameter i s included) and i s a generalization of the exponential, i t i s l i k e l y that we would obtain a somewhat better f i t . However the Weibull was not pursued for two reasons. F i r s t l y , with so much data, the improvement i n f i t would probably be s l i g h t ( i n fact any model would probably be rejected). Secondly, the subsequent i n t e r p r e t a t i o n would be d i f f i c u l t - i t would not be clear which parameter to focus on, and meaningful r e s u l t s would be hard to obtain with so many parameters and 2.4 Random E f f e c t s Model Instead of looking at these p a r t i c u l a r seventeen l i n e s , i t might be of more in t e r e s t to think of these l i n e s as a sample from an i n f i n i t e population of l i n e s . So we could think of an i n f i n i t e population of l i n e s of each type (material) within each system. For each of these subpopulations there would be a true underlying 8 value. It would be of i n t e r e s t to compare the underlying 8 values for the d i f f e r e n t "treatments" (types of material) and for the two systems, by examining A. the 8 values obtained for our sample of seventeen l i n e s . With only one netting l i n e i n each system and only one dowling l i n e i n system two i t would be d i f f i c u t to obtain an estimate of l i n e - t o - l i n e variance separately for these treatments. So i n i t i a l l y a random e f f e c t s model was developed incorporating only the oyster l i n e s . Throughout this s e c t i o n , f o r those l i n e s f o r w h i c h a = 5 , the e s t i m a t e of the v a r i a n c e A. of 8 t h a t was used was t h a t o b t a i n e d when the e x p o n e n t i a l model s t a r t i n g A at t= 5 was f i t t e d . A l t h o u g h , s t r i c t l y s p e a k i n g , the v a r i a n c e o f 8 s h o u l d have been e s t i m a t e d i n s t e a d i n the manner d e s c r i b e d i n S e c t i o n 2 . 2 , any r e s u l t i n g i n a c c u r a c i e s were l i k e l y to be n e g l i g i b l e . L e t B i and 82 be the t r u e u n d e r l y i n g p a r a m e t e r s f o r the h y p o t h e t i c a l p o p u l a t i o n of sys tem one and system two o y s t e r l i n e s r e s p e c t i v e l y . We w i s h t o t e s t H 0 : B i = 8 2 . I f HQ i s r e j e c t e d , t h i s would sugges t t h a t a s i g n i f i c a n t sys tem e f f e c t i s p r e s e n t . We have a sample of s i x l i n e s f rom each p o p u l a t i o n . L e t B^j be the t r u e 8 v a l u e A f o r the j t h l i n e i n the i t h sys tem and 8 ^ be the maximum l i k e l i h o o d 2 e s t i m a t e o f 8 ^ . = SECB^j) i s assumed t o be a known c o n s t a n t (no t u n r e a s o n a b l e c o n s i d e r i n g t h e l a r g e amount of d a t a w i t h w h i c h 8 j j was e s t i m a t e d ) . A T h e n , p r o v i d e d t h a t the model f i t s , s i n c e 8-jj i s a maximum l i k e l i h o o d e s t i m a t e based on a l a r g e s a m p l e , i t seems r e a s o n a b l e t o assume P i j | B i r N ( B l J , V i . ) . We assume f u r t h e r , t h a t the B j j a r e i n d e p e n d e n t l y d i s t r i b u t e d and t h a t 8 i ; j ~ N ( B i , a 2 ) 2 where o"j_ i s the l i n e - t o - l i n e v a r i a n c e (unknown) o f the t r u e p^j v a l u e s f o r the i n f i n i t e p o p u l a t i o n of l i n e s i n sys tem i . I t i s no t c l e a r t h a t t h i s second a s s u m p t i o n i s e n t i r e l y r e a s o n a b l e . We can o n l y p r o c e e d w h i l e b e a r i n g i n mind t h a t a s t r o n g a s s u m p t i o n i s b e i n g made. I t f o l l o w s t h a t - 38 -P i j ~ N ( p i ' ° i + V i j } N a i v e A n a l y s i s I n i t i a l l y a n a i v e a n a l y s i s was c a r r i e d out i n w h i c h two f a i r l y s t r o n g a s s u m p t i o n s were made: A 1. I t was assumed t h a t the s t a n d a r d e r r o r s of the B i j were n e g l i g i b l e 2 compared to the l i n e - t o - l i n e v a r i a n c e , i . e . , V^j « • I t seemed r e a s o n a b l e t o e x p e c t t h a t the V^j would be s m a l l s i n c e B j j was e s t i m a t e d w i t h a l a r g e amount of d a t a . By i g n o r i n g V^j we were i n e f f e c t , t r e a t i n g 8-y as known e x a c t l y . S i n c e we a r e n e g l e c t i n g the V^j an e s t i m a t e of i s p r o v i d e d by the A sample s t a n d a r d d e v i a t i o n of t h e B^-j's. The p r e v i o u s r e s u l t s y i e l d A A A A A a 2 = s 2 = 3- S ( B 1 . - 8 X ) 2 where pj = | E = 0 . 1 0 1 9 , 3 3 3 A A A A A a2 = S 2 = 5 E ( 8 2 j " P 2 ) 2 w h e r e 6 2 = 6 S P 2 j = ° ' 0 7 2 4 * We o b t a i n o^  = .026 and = . 0 1 1 . The v a l u e s of i n sys tem one range f rom .0050 t o .0130 and i n system two f rom .0034 to . 0 0 8 7 ; w h i l e 2 the V^j a r e s m a l l e r than our e s t i m a t e s of the a -p t h e y a r e not e n t i r e l y n e g l i g i b l e . 2. The second a s s u m p t i o n was t h a t the l i n e - t o - l i n e v a r i a n c e was the same f o r b o t h s y s t e m s , i . e . a\ = o^. S i n c e we have assumed t h a t the - 39 -Pij's are normally distributed we can test the hypothesis c± = 0 2 by 2 2 comparing S^  /S 2 to F 5 5 tables. o 2 •51- = 5.67 ===> p = .05. c 2  s2 So the assumption o"i = o 2 appears to be a f a i r l y strong one. Under assumptions 1. and 2. we have a sample of six 8 values from each of two normal populations with equal variance. So H Q : 8i = 82 can be tested by a two sample t-test: 6 1 " 6 2 = 2.57 ===> p = 0.015 S 4 + 7T p 6 6where S = / 5 S ^ * 5 S ^ = 0.020. p 10 This seems to suggest that there is a difference between the underlying B values for the two systems. However this analysis relied on two fa i r l y strong assumptions. The second assumption, a\ - 02 w a s tested by means of an F-test, which is particularly sensitive to our assumption that the 8 i j ' s &re normal. A more careful analysis is called for. Second A n a l y s i s A This time the standard error of the P ij's was not neglected. This analysis was done both assuming 01 = a 2 and also without this assumption. - 40 -Case 1: Assume Q\ = OY - O If we assume that the line-to-line variance is the same in both systems then there are three parameters - Bi, B2 and a to be estimated. A A 2 since 8^ NC^, a + V^)> the likelihood of obtaining the twelve B's that we obtained is L = 2 n 6 n i=l j=l /2* / o 2 + exp -1/2 Our estimate of the variance of B-jj was made with a large amount of data, so i t can be assumed to be close to the true variance V J J . So we replace V ^ j with this estimate, thus treating V ^ j as a known constant. logL Z E i j 2 ( P i i " hy log(o 2 + V ) + — i J i -i J ( a 2 + V ) + constant ologL 5 8 i = E (Pi, - Pi) j (a + Vx ) o 2logL o 6 i 2 = -E « 0) j (a + Vj ) ologL ap2 » E Cp2j " P2) j (<r + v 2 j ) 5 2logL ap2 2 = -E « 0) j (a + V 2 j) »i2Bt- - I E E i J ( P i 1 - Pl>' (o2+V±.) (a 2 + V ) 2 - 41 -To f i n d the maximum l i k e l i h o o d estimates of 61, 82 and a we need to solve simultaneously = = ^ 2Sk = 0 . ° P i o82 S o2 Let w. . = - i = 1,2. 1 J a2 + V, . I j A T h e n ^ k , „ „ . > h . i — - 1.1 . 2 . 1 • l j 3 Substituting the above values (81* and 82* s a y ) i n S l o g L , 9 a 2 2 w i i 8 i i — « 0 ===> Z Z[w - w (6. . =— ) ] = 0 and 9a 2 i j i j 1 3 1 J 2 W i j a 2 l o g L ( 8 i * , 8?*, a 2) = I s E w2 A A E w. . 8., * E w. . 8. . j ^ ^ . j ^ * J i j J J . i j J J j i j A A E w2. p±. £ w ± j - E w ± j p ± j E w2. + (J J 1 )} . (E w ) 2 3 aiogL - 2 — - — = 0 can be solved by a Newton Raphson i t e r a t i o n to obtain cr . A numerical check can be done to v e r i f y that dlogL = 0 does i n fact aa 2 2 2 y i e l d a maximum. Furthermore since ^ l°gk a m j 0 l°gL a r e both < 0, 2 2 a p i a p 2 - 42 -d ^ g L =0 and Qjjj° g L = 0 w i l l c e r t a i n l y y i e l d maxima. H a v i n g o b t a i n e d a 2 A A i t i s s t r a i g h t f o r w a r d to o b t a i n 6i and 62 as we have e x p l i c i t A 2 e x p r e s s i o n s f o r them i n terms of a . To t e s t whether the u n d e r l y i n g p a r a m e t e r s 81 and 82 a re the same, we need to know the v a r i a n c e of our e s t i m a t e s . An e s t i m a t e of the A A A 2 c o v a r i a n c e m a t r i x o f (81, P2> o ) i s p r o v i d e d by the i n v e r s e of the n e g a t i v e of the m a t r i x of second d e r i v a t i v e s of l o g L e v a l u a t e d at t h e maximum l i k e l i h o o d e s t i m a t e s . A l t e r n a t i v e l y we can p r o c e e d as f o l l o w s : A A Z w. . 6, . i j i j A 1 8 = .3 t where w . = i = 1 ,2. i ^ 12 Suppose the w^j were known. Then we would have A E w , . B. . , i j K i j * l g = and v a r 8 = •= 0 i L w. . i L w. . i = 1 ,2. i j i j A S u b s t i t u t i n g our e s t i m a t e w „ of w .^. we o b t a i n the f o l l o w i n g rough a p p r o x i m a t i o n t o the v a r i a n c e o f p^: v a r 8 ± = — j i = 1 ,2. - 43 -R e s u l t s /v A . B i = 0 . 0 9 8 3 , 8 2 = 0 . 0 7 2 2 , a = 0 . 0 1 5 8 . A A E s t i m a t e d S E ( p i ) = 0 . 0 0 7 6 , e s t i m a t e d S E ( B 2 ) - 0 . 0 0 6 8 . These l e a d t o A A PI - P2 2 > 5 6 A A A A / ( S E ( B i ) 2 + S E ( 8 2 ) 2 ) w h i c h i s a l m o s t i d e n t i c a l t o the v a l u e of the s t a t i s t i c o b t a i n e d i n the n a i v e a n a l y s i s . Compar ing the v a l u e 2.56 to normal t a b l e s would a g a i n s u g g e s t t h a t t h e r e i s a s i g n i f i c a n t d i f f e r e n c e between B i and 6 2. However t h i s t e s t i s not c o m p l e t e l y l e g i t i m a t e f o r two r e a s o n s : 2 F i r s t l y 8^  and 82 a r e not i n d e p e n d e n t s i n c e they bo th i n v o l v e a and s e c o n d l y i t would seem more r e a s o n a b l e t o compare the v a l u e of the A A s t a t i s t i c t o t - t a b l e s as the v a r i a n c e s of p± and p 2 a r e o n l y e s t i m a t e s . The p r o b l e m i s t h a t i t i s not c l e a r how many degrees of f reedom a r e a p p r o p r i a t e . I f the square of the denomina to r of our s t a t i s t i c were a l i n e a r c o m b i n a t i o n of i n d e p e n d e n t c h i - s q u a r e d v a r i a b l e s , S a t t e r t h w a i t e ' s a p p r o x i m a t i o n ( S a t t e r t h w a i t e , 1946) c o u l d be used to e s t i m a t e the degrees of f r e e d o m . A A However our e s t i m a t e s of the v a r i a n c e s of 81 and 82 a r e o f a more c o m p l i c a t e d form and S a t t e r t h w a i t e ' s a p p r o x i m a t i o n i s not a p p l i c a b l e h e r e . A l s o , t h e whole a n a l y s i s was based on our a s s u m p t i o n of n o r m a l i t y f o r the B-LJ 'S. Because of t h e s e prob lems an e x a c t p - v a l u e cannot be g i v e n . However the v a l u e of our t e s t s t a t i s t i c i s s u f f i c i e n t l y l a r g e t o - 44 -suggest a s i g n i f i c a n t difference between 61 and 82> even allowing for s l i g h t inaccuracies. C a s e 2: o~i and a? n o t assumed e q u a l If a\ and o"2 are not assumed to be equal then we need estimates of four parameters - 6 1 , 6 2 , cri, 0 2 . For the system one oyster l i n e s A A A the l i k e l i h o o d under our model of obtaining 6 n , 612,.. . ,P16 i s L = n j exp ^2TC , 2 _,_ „ • a + V i (Pi, " Pi)' -1/2 ^ (o^ + VU) 1 J and logL = - Y E , (Pi, - Pi)' In ( c r + V j . ) + J -(<T + V i . ) 1 3 + constant. If we l e t w, , then £ "1.1 "'.1 E wj . As before we can substitute i n d l o g l Then - = 0 can be solved by bai' boi' Newton Raphson to obtain a\ . Then i t i s easy to f i n d 81• Analogously A to the previous case, an estimate of the variance of 6 l Is given by A var Bj * j m j o i 2 + V L - 45 -S i m i l a r l y the l i k e l i h o o d of obtaining 6 2 1 , . . ' 8 2 6 i s 6 L = n exp i=l /2n /cr + Vo. 2 Z3 ( 6 2 , " 8 2 ) ' - 1/2 2 (cr2 + V 2.) 2 J and o"2 , 82 as well as an estimate of the variance of 82 can be obtained i n the same way. Results A A Bi = 0.0997, p 2 = 0.0719. A A ai = 0.0219, a2 = 0.0089. A A Estimated SECp^ = 0.0098, estimated SE(p 2) = 0.0043. Pi " P2 A A A A ,2 , o r , , „ N 2 -= 2.59 / ( S E ( p i )  + S E ( p 2 ) 2 ) The same problems a r i s e as for the case where we assumed ai = 02, except that t h i s time Pi and P2 aren't correlated. Again an exact p-value cannot be quoted, but again a s i g n i f i c a n t difference between Pi and P2 i s strongly suggested. Our test s t a t i s t i c w i l l be found s i g n i f i c a n t at the 5% l e v e l i f i t i s compared to t f tables for any f >_ 3. It seems to make very l i t t l e d ifference whether or not we assume o"i = o"2; the r e s u l t i s almost i d e n t i c a l for both cases. So the fact that A A for case 1, Pi and P2 are correlated probably doesn't a f f e c t the ana l y s i s too much. However the fact remains that our analysis depends on our assumption of normality for the P i j ' s an^ 0 1 3 7 he se n s i t i v e to t h i s assumption. A summary of the r e s u l t s i s given below: - 46 -E s t i m a t e ± s t a n d a r d e r r o r Second A n a l y s i s Second A n a l y s i s w i t h o u t assuming P a r a m e t e r N a i v e A n a l y s i s assuming o"i = a 2 al ~ 02 Bi 0.1019 ± .0041 0.0983 ± .0076 0.0997 ± .0098 62 0.0724 ± .0024 0.0722 ± .0068 0.0719 ± .0043 Our a n a l y s e s s u g g e s t t h a t 81 and 82 a r e s i g n i f i c a n t l y d i f f e r e n t w h i c h can be i n t e r p r e t e d as meaning t h a t sys tem i s an i m p o r t a n t f a c t o r i n d e t e r m i n i n g the s u r v i v a l d i s t r i b u t i o n , a t l e a s t of the o y s t e r l i n e s . S i g n i f i c a n t sys tem e f f e c t i s c e r t a i n l y not s u r p r i s i n g as when the n o n p a r a m e t r i c t e s t s were c a r r i e d out t h e b i g g e s t d i f f e r e n c e s o c c u r r e d when l i n e s f rom d i f f e r e n t systems were compared . 2.5 Incorporating Netting and Dowling Lines i n t o the Model I t i s not o n l y the sys tem e f f e c t w h i c h i s of i n t e r e s t . I t i s a l s o of i n t e r e s t t o know whether the t y p e of m a t e r i a l p l a y s a r o l e i n d e t e r m i n i n g the s u r v i v a l d i s t r i b u t i o n . W i t h o n l y one n e t t i n g l i n e i n e a c h sys tem and one d o w l i n g l i n e i n sys tem two i t w i l l not be p o s s i b l e to o b t a i n an e s t i m a t e of l i n e - t o - l i n e v a r i a n c e f o r each type of m a t e r i a l - 47 -within each system. It w i l l have to be assumed that l i n e - t o - l i n e 2 2 variance i s independent of type of material. As before cr^  and o"2 are the l i n e - t o - l i n e variances for systems one and two respectively (regardless of material type). The previous model i s extended to incororate netting and dowling l i n e s as follows: Let P^jk be the true 8 value for the kth l i n e of the j t h type i n system i . Let B^jk be our maximum l i k e l i h o o d estimate of B-jjk and l e t V\£jk = variance of S ^ j ^ . assumed to be a known constant. Assume ^kKjk ^i jk 'W and 8 ± j k N(8 + f± + t., o 2) where f = system e f f e c t i=l,2, and £ f = 0, i t. = type e f f e c t j=l,2,3, and E t = 0. 1 j j So t i = e f f e c t due to type 1 ( o y s t e r ) , t 2 = e f f e c t due to type 2 (dowling), t3 = e f f e c t due to type 3 ( n e t t i n g ) . We are assuming no system-type i n t e r a c t i o n . The variance of 2 2 P i j k * s °i instead of a±^ as we don't have s u f f i c i e n t r e p l i c a t i o n to allow the variance to depend on type. As before the assumption of normality f o r the 8 i j k ' s * s c e r t a i n l y reasonable, but the assumption of normality for the 8 i j k ' s ^ s P u r e ± y a n assumption and cannot be v e r i f i e d . It follows that Bijk N(8 + f i + t . , o 2 + V i j k ) . In order to compare the slopes of the s u r v i v a l curves for the two systems and for the three types of material, we need estimates of 8, f i , - 48 -2 2 f 2 , t j , t 2 , t 3 , o i and o 2 and e s t i m a t e s of the v a r i a n c e s and c o v a r i a n c e s of f j , f 2 , t i , t 2 , t 3 . The l i k e l i h o o d under our model o f A o b t a i n i n g the s e v e n t e e n P i j k ' s t h a t we o b t a i n e d i s L = n n n • exp i j k VZ% / a 2 + l i j k S u b s t i t u t i n g f 2 = -fy and t 3 = - t i - t 2 , the maximum l i k e l i h o o d 2. 2 e s t i m a t e s (01 , o 2 , f^, t^, t 2 , P) a r e o b t a i n e d by Newton Raphson . 2 2 D u r i n g the i t e r a t i o n , 0^  , and o 2 a re c o n s t r a i n e d to be n o n n e g a t i v e . R e s u l t s A 8 = 0 . 0 8 6 2 . A f x = 0 .0176 ===> f 2 = - 0 . 0 1 7 6 . A A A t i = 0 . 0 0 2 4 5 , t 2 = 0 . 0 0 1 0 5 , t 3 » - . 0 0 3 5 0 . a i = . 0 2 3 4 , 0 2 = . 0 0 7 2 . A A A A The P^jk = p + f^ + t j a r e g i v e n be low: Type System O y s t e r N e t t i n g D o w l i n g 1 0 .1063 0 .1003 0 .1049 2 0 .0711 0.0651 0.0697 Note t h a t the e s t i m a t e s of the s l o p e f o r r e s p e c t i v e l y the system one and - 49 -t h e system two o y s t e r l i n e s (0 .1063 and 0 .0711) a r e s i m i l a r but n o t i d e n t i c a l t o the e s t i m a t e s o b t a i n e d b e f o r e under t h e model f o r t h e o y s t e r l i n e s a l o n e ( r e s p e c t i v e l y 0.0997 and 0 . 0 7 1 9 ) . E s t i m a t e s o f the v a r i a n c e s o f t h e s e e s t i m a t e s a r e o b t a i n e d as u s u a l u s i n g the m a t r i x of second d e r i v a t i v e s e v a l u a t e d a t the maximum l i k e l i h o o d e s t i m a t e s , wh ich was a l r e a d y c a l c u l a t e d d u r i n g the Newton Raphson p r o c e s s . A A SE 8 = . 0 0 3 5 . A A A A SE f i = SE f 2 = . 0 0 3 1 . A A A A SE ti = . 0 0 3 0 , SE t 2 = . 0 0 3 8 , SE t 3 = . 0 0 3 9 . I t was t h e n o f i n t e r e s t to t e s t f o r d i f f e r e n c e s between f i and f 2 and between t^, t 2 and t 3 : f i ~ f; = 5.62 / v a r ( f i - f 2 ) A A t l " t 2 = 0 .25 A A A / v a r ( t i - t 2 ) A A 1 : 1 " t 3 = 1.02 A A A / v a r ( t i - t 3 ) S i n c e t i + t 2 + t 3 = 0 i t i s u n n e c e s s a r y a l s o to compare t 2 to t 3 . As b e f o r e , i t seems r e a s o n a b l e to compare t h e s e q u a n t i t i e s t o t - t a b l e s s i n c e the v a r i a n c e s a r e e s t i m a t e d . A g a i n i t i s not c l e a r how many - 50 -d e g r e e s of f reedom s h o u l d be u s e d . S i n c e we a r e d o i n g t h r e e c o m p a r i s o n s the s i g n i f i c a n c e l e v e l s h o u l d be a d j u s t e d a c c o r d i n g l y : i f the o v e r a l l s i g n i f i c a n c e l e v e l i s t o be 5%, t h e n s i g n i f i c a n c e l e v e l of e a c h c o m p a r i s o n c o u l d be 5/3% = 1.66%. F o r the above c o m p a r i s o n s , whatever the d e g r e e s of f reedom we r e a c h the s a m e c o n c l u s i o n , namely t h a t t h e r e i s a s i g n i f i c a n t d i f f e r e n c e between t h e 8 v a l u e s f o r the two sys tems (as was found p r e v i o u s l y ) , but no s i g n i f i c a n t d i f f e r e n c e between the 8 v a l u e s f o r the t h r e e t y p e s of m a t e r i a l . T h i s s u g g e s t s t h a t r e g a r d i n g the r a t e a t w h i c h the b a r n a c l e s d i e o f f , t h e r e i s a s i g n i f i c a n t d i f f e r e n c e between the two sys tems but not between the t h r e e t y p e s o f m a t e r i a l . Discussion I n s p e c t i n g t h e 8 v a l u e s , i t i s seen t h a t the v a l u e s f o r s y s t e m one a r e g e n e r a l l y l a r g e r than t h o s e f o r sys tem two, whereas p v a l u e s f o r any p a r t i c u l a r type o f m a t e r i a l show no t e n d e n c y to be b i g g e r or s m a l l e r t h a n t h o s e f o r any o t h e r t y p e . These o b s e r v a t i o n s a r e bo rne out by t h e above a n a l y s e s w h i c h s u g g e s t a s i g n i f i c a n t sys tem e f f e c t but n o t a s i g n i f i c a n t t y p e e f f e c t . I t i s seen f rom a\ and a 2 t h a t 8 v a l u e s a r e more v a r i a b l e i n sys tem one t h a n i n sys tem two. A. S t a t i s t i c a l a n a l y s e s r e l a t i n g to the a v a l u e s were not c a r r i e d b u t an i n s p e c t i o n of the v a l u e s s u g g e s t s t h a t t h e r e i s b o t h a sys tem and t y p e e f f e c t . The a v a l u e s f o r sys tem one a r e g e n e r a l l y s m a l l e r than v a l u e s f o r sys tem two and v a l u e s f o r d o w l i n g l i n e s a r e g e n e r a l l y s m a l l e r t h a n v a l u e s f o r the o t h e r t y pe s of l i n e . The p h y s i c a l i n t e r p r e t a t i o n of - 51 -these observations i s that barnacles on dowling and system one l i n e s tend to st a r t dying e a r l i e r than other barnacles. Barnacles i n system one tend to die at a f a s t e r rate than those i n system two. Furthermore the rate at which the system one barnacles die shows more v a r i a b i l i t y from l i n e to l i n e . The f i n a l observation relates to the i n i t i a l number of barnacles on the l i n e s . There are more barnacles i n i t i a l l y on a l i n e of a given type i n system two than on a l i n e of the same type i n system one. Also A A we found the 8 values to be generally smaller and the a values generally larger f o r system two l i n e s than for system one l i n e s . Formal s t a t i s t i c a l tests are not appropriate and even the simple c o r r e l a t i o n c o e f f i c i e n t i s not very meaningful with such a small number of l i n e s of a given type. However i t i s worth noting that f o r the system one oyster l i n e s , the c o r r e l a t i o n c o e f f i c i e n t between the i n i t i a l number of barnacles and 6 i s - .93. This c o r r e l a t i o n suggests that on l i n e s with fewer barnacles i n i t i a l l y , barnacles tend to st a r t dying e a r l i e r and at a fas t e r rate. However the c o r r e l a t i o n may also be due to a t h i r d variable related to both 6 and i n i t i a l number. This w i l l be discussed again l a t e r i n r e l a t i o n to the growth data. A l l i n a l l i t appears that system i s more important than type as regards s u r v i v a l d i s t r i b u t i o n . However t h i s conclusion i s somewhat tentative since i t i s based on the analysis of parameters p and a of a model which doesn't adequately f i t the data. Furthermore the subsequent analyses required a strong normality assumption. - 52 -3 . EXAMINATION OF THE GROWTH DATA 3 .1 Growth D a t a Growth d a t a was c o l l e c t e d at t = 1, 2 , . . . , 9 , 10, 11 , 14, 17 weeks . A t each of t h e s e t i m e s samples were taken as f o l l o w s : f rom each s h e l l on each o y s t e r l i n e and f rom e a c h p i e c e of wood on each d o w l i n g l i n e a h a p h a z a r d sample of 5 l i v e b a r n a c l e s was s e l e c t e d . From each p i e c e o f r u b b e r on each n e t t i n g l i n e a h a p h a z a r d sample of 10 l i v e b a r n a c l e s was s e l e c t e d . I t w i l l be assumed t h a t on each i t e m e v e r y l i v e b a r n a c l e was e q u a l l y l i k e l y t o be s a m p l e d , so f o r example the p r o b a b i l i t y of b e i n g sampled was no t r e l a t e d t o s i z e . F o r each sampled b a r n a c l e s h e l l l e n g t h and neck l e n g t h were m e a s u r e d . Measurements were r e c o r d e d i n cm. The sampled b a r n a c l e s were not removed f rom the i t e m s t o w h i c h they were a t t a c h e d . I n t o t a l t h e r e was an enormous amount of growth d a t a : the t o t a l number of b a r n a c l e s measured a t each t ime was: From o y s t e r l i n e s : 2 x 6 x 1 0 x 5 = 6 0 0 , From d o w l i n g l i n e s : 3 x 20 x 5 = 300, From n e t t i n g l i n e s : 2 x 10 x 10 = 200 . The d a t a was i n c o m p l e t e i n s e v e r a l ways: a t t = 1 week, o b s e r v a t i o n s were t a k e n o n l y on b a r n a c l e s f rom 3 o y s t e r l i n e s and 1 d o w l i n g l i n e i n sys tem 2. A l s o a t t = 10 and 11 weeks o b s e r v a t i o n s were t a k e n on b a r n a c l e s f rom some of the l i n e s but not a l l of them. I t somet imes happened t h a t fewer than f i v e ( o r t e n ) b a r n a c l e s were sampled f rom a p a r t i c u l a r i t e m . T h i s was because the p o p u l a t i o n f rom wh ich the - 53 -b a r n a c l e s were sampled was c o n t i n u a l l y g e t t i n g s m a l l e r due t o b a r n a c l e s d y i n g . Somet imes , towards the end of the s tudy ( p a r t i c u l a r l y f o r the s y s t e m one o y s t e r l i n e s where i n i t i a l sample s i z e s were s m a l l ) fewer than f i v e ( o r t e n ) l i v i n g b a r n a c l e s rema ined on a p a r t i c u l a r i t e m . I n t h i s case a l l r e m a i n i n g b a r n a c l e s were s a m p l e d . The a n a l y s e s w h i c h were used assumed t h a t measurements were t a k e n a t the same t ime p o i n t s f o r a l l i n d i v i d u a l s . So o n l y the d a t a c o l l e c t e d a t t = 2 , 3 , 4 . . . , 9 , 14, 17 weeks was u s e d . The a n a l y s e s c o u l d have been e x t e n d e d t o accommodate the m i s s i n g d a t a and methods of h a n d l i n g m i s s i n g d a t a w i l l be d i s c u s s e d l a t e r . However t h i s would have made the c o m p u t a t i o n much more l a b o r i o u s and s i n c e t h e amount of d a t a n e g l e c t e d i s r e l a t i v e l y s m a l l i t w a s n ' t c o n s i d e r e d w o r t h w h i l e . A l s o i t ems were n o t i n c l u d e d i f a t any p o i n t i n t ime fewer than f i v e ( o r t e n ) b a r n a c l e s were sampled f rom them. A g a i n not much i n f o r m a t i o n was l o s t by d o i n g t h i s . O n l y t h e d a t a r e l a t i n g t o s h e l l l e n g t h was a n a l y z e d . The d a t a r e l a t i n g to neck l e n g t h c o u l d have been t r e a t e d i n an i d e n t i c a l way. I n summary: on e a c h o f 17 l i n e s t h e r e a r e a number of i t e m s ( o y s t e r s h e l l s , p i e c e s of wood or r u b b e r ) . At each p o i n t i n t i m e , measurements a r e made on a random sample of b a r n a c l e s t a k e n f rom each i t e m . The p o p u l a t i o n f rom w h i c h t h e s e b a r n a c l e s a r e sampled i s c o n t i n u a l l y d i m i n i s h i n g because b a r n a c l e s a r e d y i n g . The p r o p o r t i o n of b a r n a c l e s sampled on a p a r t i c u l a r i t e m i s a p p r o x i m a t e l y 10% i n i t i a l l y and 20-30% towards the end of the s t u d y e x c e p t f o r i t ems on the sys tem one o y s t e r l i n e s where the p r o p o r t i o n s a r e somewhat l a r g e r - a p p r o x i m a t e l y 25% i n i t i a l l y and c l o s e t o 100% by - 54 -the end of the study. These proportions are only estimates as we don't a c t u a l l y have the s u r v i v a l data for each item, only for each l i n e . I ndividual barnacles cannot be i d e n t i f i e d . We would l i k e to f i t growth curves for the s h e l l length of the barnacles and to compare these curves for the two systems and for the three types of material. 3.2 Growth Curves f o r Items Sophisticated growth curve models have been developed f o r l o n g i t u d i n a l data - for the s i t u a t i o n where an observation i s obtained at each time point on each of a number of i n d i v i d u a l s . In our case we don't have l o n g i t u d i n a l data for the barnacles: at each point we have measurements on a sample of u n i d e n t i f i e d barnacles - each barnacle may or may not have been measured at the previous time point. The items, however, can be i d e n t i f i e d and i f we were to average over the sample obtained from each item at each time point, we could think of our data as l o n g i t u d i n a l data for the items: on each l i n e we would have a number of items and we would have a measurement representing each item at each time. This measurement would a c t u a l l y be the average s h e l l length over a sample of barnacles picked from the remaining l i v i n g barnacles on that item at that time. Since each barnacle has equal p r o b a b i l i t y of being sampled, regardless of s i z e , and since the proportion of barnacles sampled i s quite large, i t should be reasonable to allow this average to represent a l l the remaining l i v i n g barnacles on that items at that time. So - 55 -i n s t e a d of f o l l o w i n g i n d i v i d u a l b a r n a c l e s we would be f o l l o w i n g i n d i v i d u a l i t e m s . Suppose we would l i k e t o f i t p o l y n o m i a l growth c u r v e s . W i t h e f f e c t i v e l y o n l y one o b s e r v a t i o n per i t e m we w i l l have to assume t h a t a l l i t e m s w i t h i n a l i n e have t h e same p a r a m e t e r s . So c o m p a r i s o n of i t e m s w i l l be i m p o s s i b l e . But i n any case the main i n t e r e s t l i e s i n c o m p a r i s o n o f the 17 l i n e s . We may a l l o w e a c h of the 17 l i n e s t o have d i f f e r e n t p a r a m e t e r s and we may t h i n k o f the i t ems on each l i n e as a sample f rom a c o n c e p t u a l p o p u l a t i o n o f i t ems on t h a t l i n e . We may t h e n f i t the a v e r a g e growth c u r v e f o r i t e m s on each l i n e . What w i l l t h i s c u r v e r e p r e s e n t ? C o n s i d e r the growth c u r v e f o r the f i r s t o y s t e r l i n e . T h i s w i l l show how the a v e r a g e s h e l l l e n g t h of the r e m a i n i n g l i v i n g b a r n a c l e s on a t y p i c a l i t e m f rom t h i s l i n e changes w i t h t i m e . (We hope t h a t the a v e r a g e over a l l b a r n a c l e s on a p a r t i c u l a r i t e m a t a p a r t i c u l a r t ime w i l l be w e l l - r e p r e s e n t e d by the a v e r a g e over a sample of b a r n a c l e s t a k e n f rom t h a t i t e m a t t h a t t i m e ) . T h i s a v e r a g e s i z e changes f o r two r e a s o n s - f i r s t l y because the b a r n a c l e s grow and s e c o n d l y because the p o p u l a t i o n o f l i v i n g b a r n a c l e s changes due t o b a r n a c l e s d y i n g . So the growth c u r v e r e p r e s e n t s the n a t u r a l p o p u l a t i o n r a t h e r than i n d i v i d u a l b a r n a c l e s . We would e x p e c t the growth c u r v e s f o r the n a t u r a l p o p u l a t i o n t o be s i m i l a r t o the growth c u r v e s f o r i n d i v i d u a l b a r n a c l e s , p r o v i d e d t h a t the p r o b a b i l i t y of d y i n g i s u n r e l a t e d t o s i z e . I f the b a r n a c l e s t h a t d i e i n a p a r t i c u l a r t i m e - i n t e r v a l a r e n e i t h e r p a r t i c u l a r l y b i g n o r p a r t i c u l a r l y s m a l l , t h e n the change i n p o p u l a t i o n w i l l no t a f f e c t t h e a v e r a g e s i z e o f - 56 -t h e l i v i n g b a r n a c l e s . The changes i n t h i s a v e r a g e s i z e w i l l then be e n t i r e l y due to the growth of the b a r n a c l e s . I f , on the o t h e r h a n d , t h e r e i s a t e n d e n c y f o r the l a r g e r b a r n a c l e s , s a y , t o d i e f i r s t , t h e n w h i l e each i n d i v i d u a l b a r n a c l e grows b i g g e r , the average s i z e of the l i v i n g b a r n a c l e s might a c t u a l l y get s m a l l e r . I f t h i s were the c a s e , the growth c h a r a c t e r i s t i c s of the n a t u r a l p o p u l a t i o n would not r e f l e c t t h e growth c h a r a c t e r i s t i c s of i n d i v i d u a l b a r n a c l e s . W i t h o u t b e i n g a b l e to i d e n t i f y i n d i v i d u a l b a r n a c l e s i t i s d i f f i c u l t t o a s c e r t a i n whether the p r o b a b i l i t y of d y i n g i s r e l a t e d to s i z e . T h i s q u e s t i o n w i l l be a d d r e s s e d l a t e r a l o n g w i t h a d i s c u s s i o n of the l e g i t i m a c y of t h e a p p r o a c h and of the a s s u m p t i o n s t h a t a r e i m p l i c i t l y b e i n g made. So u s i n g the " l o n g i t u d i n a l " d a t a f o r i t e m s , p o l y n o m i a l g rowth c u r v e s were f i t t e d f o r each of the 17 l i n e s . Two growth c u r v e models were employed - the f i r s t was due t o P o t h o f f and Roy (1964) and the second t o Rao (1965 , 1 9 6 6 ) . The c u r v e s o b t a i n e d f o r the two d i f f e r e n t systems and f o r the t h r e e t y p e s of m a t e r i a l were t h e n compared . 3.3 The Growth Curve Models Pothoff and Roy's Model I t i s assumed t h a t l o n g i t u d i n a l d a t a i s a v a i l a b l e - s u c c e s s i v e measurements on a number of i n d i v i d u a l s . The measurements f o r any one i n d i v i d u a l a r e t h e n c l e a r l y c o r r e l a t e d . T y p i c a l l y the i n d i v i d u a l s a r e d i v i d e d i n t o groups w h i c h a r e to be compared . I t i s assumed t h a t each i n d i v i d u a l has measurements t a k e n a t the - 57 -same q points i n time. The set of q measurements for any one i n d i v i d u a l form one row of the data matrix Yo. The model i s as follows: E[Y 0] = ' A I P nxq nxm mxp pxq where n = number of i n d i v i d u a l s , m = number of groups, q = number of time points, p = number of parameters f i t t e d within each group (p _< q), and A i s the design matrix across i n d i v i d u a l s , £ i s a matrix of parameters to be estimated, P i s a matrix of known constants related to time (assumed to be of f u l l rank, p). It i s assumed that the rows of YQ are mutually independent (measurements on d i f f e r e n t i n d i v i d u a l s are independent) and that the q elements i n any one row follow the multivariate normal d i s t r i b u t i o n with unknown covariance matrix T.Q (q x q), the same for every i n d i v i d u a l . So i s not diagonal since successive measurements on the same i n d i v i d u a l are correlated. It would be possible to extend the model to allow the covariance matrices for d i f f e r e n t i n d i v i d u a l s to be multiples of each other and also to allow d i f f e r e n t i n d i v i d u a l s to have measurements taken at d i f f e r e n t times. These extensions are appropriate i n our s i t u a t i o n and w i l l be considered l a t e r . To i l l u s t r a t e the model consider the following s i t u a t i o n : we have m groups of i n d i v i d u a l s with nj i n the j t h group. The growth curves for i n d i v i d u a l s i n a l l groups are from the same family, for example - 58 -p o l y n o m i a l s of degree p-1. But the p a r a m e t e r s may be d i f f e r e n t f o r the d i f f e r e n t g r o u p s . So the growth c u r v e a s s o c i a t e d w i t h the j t h group i s Then A = { n { n { m • 0 0 0 1 0 1 '10 '1 p-1 m^0 ^m p-1 P = .P-1 .P-1 A l t e r n a t i v e l y i f the t ime p o i n t s a r e e q u a l l y s p a c e d , o r t h o g o n a l c o e f f i c i e n t s may be used f o r P. T h i s can be g e n e r a l i s e d t o the s i t u a t i o n where t h e r e a r e two or more f a c t o r s each w i t h a number of l e v e l s or t o the m u l t i - r e s p o n s e - 59 -s i t u a t i o n where more t h a n one growth c h a r a c t e r i s t i c i s measured . Any of the u s u a l f a c t o r i a l or b l o c k d e s i g n s can be accommodated by t h i s m o d e l . We want to e s t i m a t e the p a r a m e t e r s E,^ and to t e s t v a r i o u s h y p o t h e s e s of the form C I V = 0. s x m m x p p x u F o r example , i f the h y p o t h e s i s i s t h a t a l l m growth c u r v e s a re the same, i . e . t h a t a l l p a r a m e t e r s f o r a l l g roups a r e e q u a l , t h e n C - ( I -1 ) and V = I . ( m - l ) x ( m - l ) ( m - l ) x l pxp I f the h y p o t h e s i s i s t h a t a l l c u r v e s a r e of degree p-2 or l e s s t h e n C = I and V mxm ( p x l ) S o l u t i o n P o t h o f f and Roy s u g g e s t a t r a n s f o r m a t i o n w h i c h r e d u c e s the model t o a more s t a n d a r d model w h i c h has been t r e a t e d e x t e n s i v e l y i n the l i t e r a t u r e . The t r a n s f o r m a t i o n i s X = Y 0 G - 1 P 1 (P G - 1 P ' ) - 1 nxp nxq qxq qxp pxp where G i s an a r b i t r a r y symmetr i c p o s i t i v e d e f i n i t e m a t r i x such t h a t P G - 1 P» i s of f u l l r a n k . Then X i s such t h a t : ( a ) i t s rows a r e m u t u a l l y i n d e p e n d e n t , (b) t h e p e l e m e n t s i n any one row f o l l o w the m u l t i v a r i a t e normal d i s t r i b u t i o n w i t h unknown p o s i t i v e d e f i n i t e c o v a r i a n c e m a t r i x - 60 -E = [P(G')- 1 P ' ] " 1 P ( G ' ) - 1 E 0 G"1 P' (P G"1 P » ) _ 1 , pxp (c) E[X] = A?. This i s the usual multivariate analysis of variance (MANOVA) model. Under t h i s model a test of C£V = 0 i s based on S^(u x u) and S e ( u x u) where S h = (C I V)' [CjCA* A j ) - ^ ' ] - 1 (C I V) = V X' A J ( A J A ^ - 1 C J I C J C A J A j ) - 1 C | ] _ 1 CjCA' A ^ - 1 A^ X and S g = V X'[I - A^Aj^ A ^ - 1 AJ] X V, where £ i s the le a s t squares estimate of £, A^ = f i r s t r columns of A, = f i r s t r columns of C, and rank A = r ( _< m, < n) , rank C = s ( <m, <_ r) , rank V = u (< p). Several possible tests are a v a i l a b l e : (1) Roy's test i s based on the largest c h a r a c t e r i s t i c root of S, S - 1 n e (2) Hot e l l i n g ' s test i s based on the trace of S, S - 1 , h e | Se (3) Wilk's test i s based on the r a t i o of determinants S h + S e In order to test a hypothesis C£V = 0 under the o r i g i n a l model, X = Y 0 G - 1 p' (p G - 1 P')" 1 i s substituted i n t o the expressions for S h and S e. In a l l analyses which follow, Roy's test was used. This proceeds as follows: - 61 -Let X = largest c h a r a c t e r i s t i c root of S h S "~ . Compare ^ + to Heck tables (e.g. Morrison, 1976) with parameters s* = min (s, u), m* = 1/2 (|s-u|-l), n* = 1/2 (n-r-u-1). Reject the n u l l hypothesis If . ^ .. i s too large. 1 T A. (n* + 1 ) If s* = 1, the s t a t i s t i c X can be compared to F tables with (m* + 1) (2m* + 2) and (2n* + 2) degrees of freedom. The least squares estimator of £ i s given by I = (AJ A x ) _ 1 A* X = (AJ A j ) " 1 A^ Y q G _ 1 P•(P G _ 1 P ' ) _ 1 . The best l i n e a r unbiased estimator of the function C£V i s then C£V. Confidence bounds can also be found for these functions. Choice of G The analysis i s v a l i d for any G s a t i s f y i n g the given conditions. However the choice of G a f f e c t s the power of the tests and the variance of the estimators. The minimum variance unbiased estimator of £ i s (A! A . ) - 1 A! Y E - 1 P' (P S - 1 P ' ) - 1 . Comparing t h i s with the l e a s t l l l o o o squares estimator of £ given above, would suggest that the optimal choice of G i s G = E Q . Pothoff and Roy suggest that the more G d i f f e r s from E Q the worse the power of the test w i l l be and the greater the variance of the estimators. However E 0 i s unknown and an estimate of E 0 obtained from Y Q may not be used. If the - 62 -e x p e r i m e n t e r has no i d e a about the form of E Q the s i m p l e s t p r o c e d u r e i s t o use G=I. I f an e s t i m a t e of E 0 can be o b t a i n e d f rom a s i m i l a r b u t i n d e p e n d e n t exper iment i t would be p r e f e r a b l e t o use t h i s e s t i m a t e . A l t e r n a t i v e l y , i f a guess of E Q can be made b e f o r e the e x p e r i m e n t i s r u n , t h i s may be u s e d . F o r examp le , i t might be assumed t h a t the c o r r e l a t i o n between any two o b s e r v a t i o n s d p e r i o d s a p a r t i s and t h a t the v a r i a n c e i s c o n s t a n t w i t h r e s p e c t t o t i m e . Then E Q i s p r o p o r t i o n a l t o ~ i 2 3 i p p p . . . . p 1 p p . . . . 2 , p p 1 p . . . . 3 2 , p p p I . . . . . . . . . . . . U n f o r t u n a t e l y , however , the c h o i c e of p i s a r b i t r a r y u n l e s s an e s t i m a t e of i t can be o b t a i n e d f rom an i n d e p e n d e n t e x p e r i m e n t . R a o ' s M o d e l The p r o b l e m w i t h P o t h o f f and R o y ' s model i s t h a t G must be chosen a r b i t r a r i l y . Rao p o i n t s out t h a t t h e i r model d o e s n ' t u t i l i z e a l l t h e a v a i l a b l e i n f o r m a t i o n u n l e s s G = E Q and t h e y do not a l l o w an e s t i m a t e o f E Q o b t a i n e d f rom the d a t a t o be u s e d . So Rao s u g g e s t s an a l t e r n a t i v e way of r e d u c i n g the growth model t o the u s u a l MANOVA m o d e l : c o n s t r u c t a q x q n o n s i n g u l a r m a t r i x H = (H. H„) s u c h t h a t the columns - 63 -o f B-i f o rm a b a s i s f o r the v e c t o r space spanned by the rows of P and PH} = I , PH2 = 0. When rank P = p we can choose H x = G " 1 P ' ( P G - 1 P * ) - 1 , H 2 = I - P where G i s an a r b i t r a r y p o s i t i v e d e f i n i t e m a t r i x . The c h o i c e of G does no t a f f e c t e s t i m a t e s or t e s t s . L e t X = Y H = Y G " 1 P' ( P G - 1 P * ) - 1 and Z = Y H„ . o 1 o , , N o 2 nxp n x ( q - p ) T h e n , as f o r the P o t h o f f and R o y ' s m o d e l , E [X] = A£, but i n a d d i t i o n E [ Z ] =0. Hence E[x|z] = A£ + ZT where Z i s a m a t r i x of q-p c o v a r i a b l e s and T i s a m a t r i x o f unknown r e g r e s s i o n c o e f f i c i e n t s . ( q - p ) x p Rao c l a i m s t h a t the e s t i m a t e o f £ o b t a i n e d under t h i s model i s more e f f i c i e n t than t h a t o b t a i n e d under P o t h o f f and R o y ' s model as i t uses i n f o r m a t i o n c o n t a i n e d i n t h e c o v a r i a b l e s Z , w h i c h i s n e g l e c t e d under P o t h o f f and R o y ' s m o d e l . However t h i s i s not t r u e . The b e s t l i n e a r u n b i a s e d e s t i m a t o r of E, under R a o ' s model i s I = (A{ A ^ - 1 A J Y q S - 1 P' (P S " 1 P ' ) " 1 where S i s an e s t i m a t e o f E 0 o b t a i n e d f rom the d a t a : s = Y ; [ I - A i ( A - V 1 A'] Y O T h i s i s p r e c i s e l y what we would o b t a i n under P o t h o f f and R o y ' s model i f we s e t G = S, thus a l l o w i n g an e s t i m a t e of E Q o b t a i n e d f rom the d a t a as Rao d o e s . To t e s t h y p o t h e s e s of the fo rm C£V = 0, m a t r i c e s S h and S g a r e a g a i n f o u n d . T h i s t ime t h e y have a s l i g h t l y d i f f e r e n t f o r m : L e t B = [ S _ 1 - S - 1 P ' ( P S - 1 P ' ) " 1 P S - 1 ] . - 64 -Then S h - V X' A1 (Aj A ^ - 1 C ^ C j R C p - 1 (A^ A ^ - 1 A{ X V S = V (P S " 1 P ' ) - 1 V e where R = (A.'A. ) - 1 + ( A . ' A . ) - 1 A.' Y B Y ' A. (A.' A . ) - 1 1 1 1 1 l o o l l l and X = Y q S " 1 P ' ( P S " 1 P ' ) " 1 . The e x p r e s s i o n f o r S e i s i d e n t i c a l t o the e x p r e s s i o n f o r S g i n P o t h o f f and R o y ' s model w i t h G s e t e q u a l t o S. However i s not the same. The t e s t s based on and S e a r e the same as f o r P o t h o f f and R o y ' s m o d e l . However when compar ing . ^ .. t o Heck t a b l e s the I i A. paramete r n* i s now s l i g h t l y d i f f e r e n t , n * = n - s - u - l - ( q - p ) . Sometimes as d i s c u s s e d by G r i z z l e and A l l e n (1969) i t i s p r e f e r a b l e t o use fewer than q-p c o v a r i a b l e s . 3.4 A p p l i c a t i o n o f t h e M o d e l s The two growth c u r v e models were a p p l i e d t o the l o n g i t u d i n a l d a t a f o r i t e m s . E a c h row of the d a t a m a t r i x c o n t a i n e d n i n e e n t r i e s . The i t h e n t r y was the a v e r a g e s h e l l l e n g t h over a random sample of l i v i n g b a r n a c l e s t a k e n a t t = f rom the i t e m c o r r e s p o n d i n g to t h a t row. The i t e m s were d i v i d e d i n t o g roups a c c o r d i n g t o the l i n e t h a t they were o n . I n t h e i r p r e s e n t form the models r e q u i r e the f o l l o w i n g a s s u m p t i o n s : (1) The rows of the d a t a m a t r i x a r e m u t u a l l y i n d e p e n d e n t . (2) The s e t of e n t r i e s i n any row f o l l o w s the m u l t i v a r i a t e n o r m a l d i s t r i b u t i o n w i t h the same c o v a r i a n c e m a t r i x E n f o r e a c h row. - 65 -Provided that we are prepared to assume that observations on barnacles from d i f f e r e n t items are independent, then the rows of our data matrix w i l l c e r t a i n l y be mutually indepndent since each row of our data matrix corresponds to a d i f f e r e n t item and no two items have any barnacles i n common. Furthermore, i f we are prepared to assume that the set of observations on an i n d i v i d u a l barnacle i s multivariate normal then the f i r s t part of assumption 2 would also be s a t i s f i e d since each entry i n our data matrix would then be the average of a number of variables assumed to be normally d i s t r i b u t e d . Whether or not i t i s reasonable to assume that the covariance matrix i s the same for every row of our data matrix w i l l be discussed l a t e r . Subsequently another analysis was c a r r i e d out i n which the weaker and more reasonable assumption was made that a set of observations corresponding to items from oyster or dowling l i n e s had covariance matrix -jr E and a set of observations corresponding to items from netting l i n e s had covariance matrix jfi ^ f ° r some f i x e d , unknown £. In order to do t h i s the model had to be modified s l i g h t l y as described i n Section 3.8. It i s not clear that even this weaker assumption i s reasonable. This too w i l l be discussed l a t e r . Only those items for which every observation was an average over f i v e (ten) barnacles were included i n the a n a l y s i s . In t o t a l 141 items were used i n the analyses. The number of items used from each l i n e was as follows: - 66 -01 02 03 04 05 06 5 8 5 6 6 8 D l D2 N l 07 08 09 11 11 8 9 8 8 010 O i l 012 D3 N2 6 10 9 13 10 The main o b j e c t i v e was to f i t p o l y n o m i a l growth c u r v e s t o each o f the l i n e s and to d e t e r m i n e whether d i f f e r e n c e s e x i s t e d between t h e s e c u r v e s f o r the t h r e e d i f f e r e n t t y p e s o f m a t e r i a l s and f o r the two s y s t e m s . As a f i r s t s t e p i t was of i n t e r e s t t o d e t e r m i n e whether l i n e s w i t h i n e a c h s y s t e m - t y p e c o m b i n a t i o n c o u l d be i g n o r e d . I f t h i s were the case i t would be r e a s o n a b l e to c o l l a p s e over l i n e s of the same type w i t h i n t h e same sys tem and we would have a two f a c t o r s i t u a t i o n w i t h the f o l l o w i n g number of i t e m s . System 1 System 2 Type O y s t e r 38 50 D o w l i n g 22 13 N e t t i n g 8 10 I n o r d e r to d e t e r m i n e whether d i f f e r e n c e s e x i s t e d between the growth c u r v e s f o r the s i x o y s t e r l i n e s i n sys tem one , p o l y n o m i a l g rowth c u r v e s were f i t t e d t o the s i x l i n e s and the p a r a m e t e r s were compared . A s i m i l a r a n a l y s i s was c a r r i e d out f o r the s i x o y s t e r l i n e s i n s y s t e m two. - 67 -3 . 5 System One Oyster Lines The a n a l y s e s were c a r r i e d out u s i n g : (a) R a o ' s m o d e l , (b) P o t h o f f and R o y ' s model w i t h G = I , ( c ) P o t h o f f and R o y ' s model w i t h G = an i n d e p e n d e n t e s t i m a t e of £ Q namely an e s t i m a t e o f E Q o b t a i n e d f rom the system 2 o y s t e r l i n e s , G - Y ' 2 [ I - AjCA} A p " 1 A;] Y Q 2 = H where Y Q 2 and A a r e r e s p e c t i v e l y the d a t a m a t r i x and t h e d e s i g n m a t r i x f o r the sys tem two o y s t e r l i n e s . F o r a l l models Y Q (38x9) = d a t a m a t r i x and ( j , k ) t h e lement of P = ( t f c - t ) J _ 1 where t = Z t = 58/9 ( s i n c e the t ime p o i n t s a r e not e q u a l l y spaced we cannot use o r t h o g o n a l c o e f f i c i e n t s ) . So P = 1 1 1 1 1 1 1 1 1 -40 -31 -22 -13 -4 5 14 23 68 9 • 9 • 9 • 9 • 9 • 9 • 9 • 9 • 9 • • • • • • • • • • * • • • • • • • • ,-40xp-l f 68 .p - l - 68 -where p i s the number p a r a m e t e r s f i t t e d . S i n c e we have t a k e n the ( j , k ) t h e lement of P to be ( t ^ - t ) 3 * r a t h e r than t^ the growth c u r v e f o r the j t h group w i l l be £ Q j + ( t - t ) +... + V i j ( t " ° p-1 The i t ems a r e d i v i d e d i n t o s i x groups w i t h the f o l l o w i n g number i n e a c h g r o u p : 5, 8 , 5, 6, 6, 8. So A (38x6) = 5{ 8{ 8{ 0 . . . . 0 0 1 . . . 0 . . . 0 1 . . . . 0 0 . . . 0 1 0 0 . . . 0 1 n = 38 (number of i n d i v i d u a l s ) , m = 6 (number of g r o u p s ) , q = 9 , r = 6. I n S e c t i o n 3 . 3 , i t was d e s c r i b e d how t o t e s t h y p o t h e s e s of the fo rm C£V = 0 under each m o d e l . By c h o o s i n g C and V a p p r o p r i a t e l y the f o l l o w i n g a n a l y s e s were c a r r i e d out under each m o d e l : 1. I t was d e t e r m i n e d what degree of p o l y n o m i a l a d e q u a t e l y f i t t e d t h e d a t a . T h i s was done i n the f o l l o w i n g way: i n i t i a l l y c u b i c s were f i t t e d and the h y p o t h e s i s t h a t a l l c u r v e s were of degree 2 or l e s s was t e s t e d . - 69 -T h i s r e q u i r e d a t e s t of £13 = Z23 ~ ••• ~ Zs3 = 0 w h i c h can be w r i t t e n C£V = 0 w i t h C = I ( 6 x 6 ) and V - 0 0 0 1 s = 6, u = 1. I f q u a d r a t i c s were found to be a d e q u a t e , q u a d r a t i c s were then f i t t e d i n o r d e r to d e t e r m i n e whether l i n e a r c u r v e s would a l s o be a d e q u a t e . I f q u a d r a t i c s were no t found t o be a d e q u a t e , q u a r t i c s were then f i t t e d t o d e t e r m i n e whether c u b i c s would be a d e q u a t e . T h i s p r o c e s s c o n t i n u e d u n t i l the a p p r o p r i a t e p o l y n o m i a l was d e t e r m i n e d . 2. The h y p o t h e s i s t h a t a l l s i x growth c u r v e s were i d e n t i c a l was t e s t e d . T h i s h y p o t h e s i s can be w r i t t e n ?10 = £20 = ••• = ?60 £ll = £21 = ••• = ^61 ^1 p-1 " ^2 p-1 " *•* " ^6 p-1 or C£V = 0 where C (5x6) 1 0 0 0 0 0 1 0 0 0 0 0 1 0 0 0 0 0 1 0 0 0 0 0 1 - 70 -and V = I ( p x p ) , s = 5, u = p, where p-1 i s the degree of the p o l y n o m i a l found to be a p p r o p r i a t e i n 1. 3 . The h y p o t h e s i s t h a t the c u r v e s were i d e n t i c a l e x c e p t p o s s i b l y f o r an a d d i t i v e c o n s t a n t ( i . e . p a r a l l e l ) was t e s t e d . T h i s h y p o t h e s i s can be w r i t t e n C£V = 0 where C i s as i n 2. and V = 0 ( p - 1 ) x ( p - 1 ) , s = 5, u = p - 1 . R e s u l t s 1. ( i ) C u b i c s were f i t t e d and the h y p o t h e s i s t h a t a l l c u r v e s were o f degree 2 or l e s s was t e s t e d . a . P o t h o f f and R o y ' s model w i t h G = Z o S h S _ 1 = . 1 1 5 , r e f e r 1.04 t o F 6 j 5 1 + , p > .1 b . R a o ' s mode l S h S e - 1 = . 1 3 8 , r e f e r 1.12 t o F 6 > l t 9 , p > .1 A c c o r d i n g to b o t h m o d e l s , q u a d r a t i c s a r e a d e q u a t e . I t i s p o s s i b l e t h a t l i n e a r c u r v e s a r e a l s o a d e q u a t e . ( i i ) Q u a d r a t i c s were f i t t e d and the h y p o t h e s i s t h a t a l l c u r v e s were l i n e a r was t e s t e d . a . P o t h o f f and R o y ' s w i t h G = I -1 S h S ~"1 = 1 .84 , r e f e r 16.6 t o T6,5h> p < .001 - 71 -b. P o t h o f f and R o y ' s model w i t h G = E J o S h S e - 1 = 1 .41 , r e f e r 12.7 t o F 6 > 5 4 , P < .001 c . R a o ' s model S h S _ 1 = 1 .86, r e f e r 14.3 t o F 6 > t | 6 , p < .001 C l e a r l y l i n e a r c u r v e s a re not adequate and q u a d r a t i c s a r e a p p r o p r i a t e . So f o r each l i n e we have e s t i m a t e s of t h r e e p a r a m e t e r s , w h i c h a r e as f o l l o w s : a . P o t h o f f and Roy , G = I 2.00 .113 - .0043 1.95 .119 - .0050 1.93 .127 - .0063 1.99 .117 - .0054 2.05 .106 - .0069 2 .00 .112 - .0049 b. P o t h o f f and Roy , G = E Q 2.00 .114 - .0042 1.95 .116 - .0047 1.91 .123 - .0057 1.98 .120 - .0051 2 .09 .102 - .0066 2.01 .107 - .0044 c . Rao 1.97 .112 - .0043 1.92 .119 - .0046 1.94 .132 - .0074 2.02 .126 - .0070 2.02 .104 - .0060 1.99 .114 - .0045 - 72 -A l l t h r e e models g i v e v e r y s i m i l a r e s t i m a t e s of £. In p a r t i c u l a r , the c h o i c e of G i n P o t h o f f and R o y ' s model a f f e c t s the e s t i m a t e v e r y l i t t l e . So , l e t t i n g y ( t ) denote s h e l l l e n g t h at t ime t , the growth c u r v e s t h a t we o b t a i n u n d e r , s a y , R a o ' s model f o r 0 1 , 0 2 , . . . , 06 a r e r e s p e c t i v e l y : y ( t ) = - .0043 t 2 + .167 t + 1.07 y ( t ) = - .0046 t 2 + .178 t + .96 y ( t ) = - .0074 t 2 + .227 t + .78 y ( t ) - - .0070 t 2 + .216 t + .92 y ( t ) = - .0060 t 2 + .181 t + 1.10 y ( t ) = - .0045 t 2 + .172 t + 1.07 A c c o r d i n g t o R a o ' s m o d e l , e s t i m a t e s of the a v e r a g e s h e l l l e n g t h over a sample of l i v i n g b a r n a c l e s t a k e n f rom an i t e m on 01 a t t = 2, 3 , 4 , 5, 6, 7, 8 , 9 , 14 weeks a r e r e s p e c t i v e l y : 1 .39 , 1 .53 , 1 .67 , 1 .80 , 1 .92 , 2 . 0 3 , 2 . 1 3 , 2 . 2 3 , 2.57 cm. P i c t u r e s of the s i x c u r v e s o b t a i n e d under R a o ' s model a r e g i v e n i n F i g u r e 1. The s i x c u r v e s a r e not t o o f a r f rom b e i n g l i n e a r and a r e v e r y s i m i l a r t o each o t h e r - a lmost p a r a l l e l , a l t h o u g h no t q u i t e i d e n t i c a l . The h y p o t h e s e s t h a t the c u r v e s were i d e n t i c a l and t h a t they were i d e n t i c a l e x c e p t f o r an a d d i t i v e c o n s t a n t were then t e s t e d w i t h the f o l l o w i n g r e s u l t s : 2. The h y p o t h e s i s t h a t a l l s i x c u r v e s were i d e n t i c a l was t e s t e d , a . P o t h o f f and R o y ' s m o d e l , G = I Figure 1. Growth Curves for the System 1 Oyster Lines 1 1 1 1 1 1 1 1 1 r — i 1 1 r .0 2.0 4 .0 6 .0 8.0 10.0 12.0 T I M E I V E E K 5 ) Figure 2. Growth Curves for the System 2 Oyster Lines - 74 -L a r g e s t r o o t o f S e _ 1 = .419 R e f e r = .29 to Heck t a b l e s w i t h s* = 3 , m* = 1/2, n* = 2 5 . P * , .04 b. P o t h o f f and R o y ' s m o d e l , G = E o L a r g e s t r o o t of S - 1 = .613 h e 613 R e f e r ,* _ = .38 to Heck t a b l e s 1 • D l J w i t h s* = 3, m* = 1/2, n* » 2 5 . p = .'01 c . R a o ' s model L a r g e s t r o o t o f S, S - 1 = .452 h e 452 R e f e r ^*. c o - = .31 to Heck t a b l e s 1.452 w i t h s* = 3 , m* = 1/2, n* = 2 5 . p « .05 We have a s u f f i c i e n t amount of d a t a to r e j e c t the h y p o t h e s i s t h a t the s i x growth c u r v e s a r e i d e n t i c a l , a l t h o u g h t h e y appear v e r y s i m i l a r i n F i g u r e 1. T h e r e i s a c o n s i d e r a b l e d i s c r e p a n c y between the p - v a l u e s o b t a i n e d under P o t h o f f and R o y ' s model w i t h G = I , and w i t h G = T.Q. I t seems t h a t the c h o i c e of G g r e a t l y a f f e c t s the v a r i a n c e of the e s t i m a t o r s and hence the p - v a l u e s . - 75 -3 . The h y p o t h e s i s t h a t t h e c u r v e s were i d e n t i c a l e x c e p t f o r an a d d i t i v e c o n s t a n t was t e s t e d , a . P o t h o f f and R o y ' s m o d e l , G = I L a r g e s t r o o t of S e _ 1 = .164 .164 R e f e r = .14 t o Heck t a b l e s 1.164 p » .05 b. P o t h o f f and R o y ' s m o d e l , G = ZQ L a r g e s t r o o t of S h S e _ 1 = .224 .224 R e f e r ^ * ~ * ^ t o H e c ^ t a b l e s p » .05 c . R a o ' s model L a r g e s t r o o t of S, S - 1 = .212 n e .212 R e f e r ^ * 212 = t o ^ e c ^ t a ° l e s p » .05 Not s u r p r i s i n g l y the c u r v e s a r e found to be p a r a l l e l . T h i s c o n c l u s i o n i s c l e a r i n s p i t e of t h e c o n s i d e r a b l e d i s c r e p a n c y w h i c h i s a g a i n o b s e r v e d between the p - v a l u e s o b t a i n e d under P o t h o f f and R o y ' s model w i t h G = S q and w i t h G = I . B e f o r e d i s c u s s i n g the r e s u l t s f u r t h e r , s i m i l a r r e s u l t s w i l l now be o b t a i n e d f o r the sys tem two o y s t e r l i n e s . - 76 -3.6 System Two Oyster Lines The same analyses were repeated for the system two oyster l i n e s . We had a new data matrix Y(50 x 9) and a new design matrix A (50x6) - 9{ 9{ 1 0 . . . . 0 . . . 0 . . . 0 1 0 . . . . 0 1 while P was unchanged. Again the analyses were carried out using Rao's model and using Pothoff and Roy's model both with G = I and with G = an estimate of E Q obtained from the system one oyster l i n e s , E = Y ' [ I - A.(A! A . ) - 1 AJ]Y o o i l l l o where Y q and A are respectively the data matrix and the design matrix for the system one oyster l i n e s . Results 1. ( i ) Cubics were f i t t e d and the hypothesis that a l l curves were of degree 2 or less was tested. a. Pothoff and Roy's model with G = E J o S h S e _ 1 = .066, re f e r .59 to F 6 > 5 1 t , p > .1 b. Rao's model S h S g 1 = .196, refer 1.6 to F g ^ g , p > .1 - 77 -C l e a r l y q u a d r a t i c s a r e a d e q u a t e . The c o n c l u s i o n i s c l e a r d e s p i t e the d i s c r e p a n c y between the two v a l u e s of the t e s t s t a t i s t i c . ( i i ) Q u a d r a t i c s were t h e n f i t t e d and the hypothe s i s t h a t a l l c u r v e s were l i n e a r was t e s t e d . a . P o t h o f f and Roy , G = I S h S e - 1 e 8 . 3 7 , r e f e r 7 5 . 3 t o F 6 > 5 1 | , P « .001 A b. P o t h o f f and Roy G = E o S h S _ 1 = 4 . 5 1 , r e f e r 4 0 . 5 t o F 6 ) 5 1 t , p « .001 c . R a o ' s model S h S e - 1 = 8 . 2 6 , r e f e r 6 2 . 9 t o F 6 > l t 6 , p « .001 The h y p o t h e s i s t h a t l i n e a r c u r v e s a r e adequate i s v e r y s t r o n g l y r e j e c t e d . So q u a d r a t i c s a r e a g a i n a p p r o p r i a t e . The e s t i m a t e s of % a r e : a . P o t h o f f and R o y , G = I 2 .23 .094 - .0052 2 .15 .104 - .0070 2 .28 .076 - .0053 2.17 .125 - .0099 2 .28 .097 - .0063 2.16 .106 - .0082 b. P o t h o f f and Roy , G = £ 2.21 .091 - .0042 2.16 .101 - .0069 2.29 .077 - .0059 2.18 .128 - .0105 2.27 .095 - .0063 2.14 .103 - .0080 - 78 -c . Rao 2.24 .094 - .0054 2.16 .107 - .0074 2.25 .081 - .0055 2.17 .122 - .0096 2.28 .099 - .0064 2.15 .107 - .0084 A g a i n , the e s t i m a t e s o b t a i n e d from a l l t h r e e methods a re a l m o s t i d e n t i c a l . The g rowth c u r v e s t h a t we o b t a i n under R a o ' s model f o r 0 7 , 08 012 a r e r e s p e c t i v e l y : y ( t ) = - .0054 t 2 + .164 t + 1.41 y ( t ) = - .0074 t 2 + .202 t + 1.16 y ( t ) = - .0055 t 2 + .152 t + 1.50 y ( t ) = - .0096 t 2 + .246 t + 0.98 y ( t ) = - .0064 t 2 + .181 t + 1.38 y ( t ) = - .0084 t 2 + .215 t + 1.11 A c c o r d i n g to R a o ' s m o d e l , e s t i m a t e s of the a v e r a g e s h e l l l e n g t h over a sample o f l i v i n g b a r n a c l e s t a k e n f rom an i t e m on 07 a t t = 2, 3 , 4 , 5, 6, 7, 8, 9 , 14 weeks a r e r e s p e c t i v e l y : 1 .72, 1 .85 , 1 .98, 2 . 1 0 , 2 . 2 0 , 2 . 2 9 , 2 . 3 8 , 2 . 4 5 , 2.65 cm. P i c t u r e s of the s i x c u r v e s o b t a i n e d under R a o ' s model a r e g i v e n i n F i g u r e 2. The c u r v e s a r e l e s s l i n e a r than t h o s e o b t a i n e d f o r the sys tem one o y s t e r l i n e s ( t h e h y p o t h e s i s o f l i n e a r i t y was r e j e c t e d w i t h a much s m a l l e r p - v a l u e f o r the sys tem two o y s t e r l i n e s than f o r the sys tem one o y s t e r l i n e s ) . They are a l s o l e s s homogeneous than the c u r v e s o b t a i n e d f o r the sys tem one o y s t e r l i n e s -whereas the c u r v e s f o r the sys tem one o y s t e r l i n e s were found to be p a r a l l e l , t h e s e c u r v e s a re f a r f rom b e i n g p a r a l l e l . T h i s o b s e r v a t i o n was c o n f i r m e d by the r e s u l t s o f the f o l l o w i n g t e s t s : 2. The h y p o t h e s i s t h a t a l l s i x c u r v e s were i d e n t i c a l was t e s t e d . a . P o t h o f f and R o y ' s m o d e l , G = I L a r g e s t r o o t of S L S _ 1 = .801 h e 801 R e f e r ^ 'gQ^ = «44 t o Heck t a b l e s 1% c r i t i c a l v a l u e = . 3 2 , p < .01 At b. P o t h o f f and R o y ' s m o d e l , G = E o L a r g e s t r o o t of S , S - 1 = .894 h e R e f e r = .47 t o Heck t a b l e s , 1% c r i t i c a l v a l u e = . 3 2 , 1% c r i t i c a l v a l u e = . 3 2 , p < .01 c . R a o ' s model L a r g e s t r o o t of S, S - 1 = .660 ° h e R e f e r ^ * = .40 to Heck t a b l e s 1% c r i t i c a l v a l u e = . 3 6 , p < .01 As e x p e c t e d the c u r v e s a r e no t found to be i d e n t i c a l . 3 . The h y p o t h e s i s t h a t the c u r v e s were p a r a l l e l was then t e s t e d . a . P o t h o f f and R o y ' s m o d e l , G = I L a r g e s t r o o t of S, S - 1 = .648 h e - 80 -R e f e r = .39 to Heck t a b l e s 1% c r i t i c a l v a l u e = . 3 0 , p < .01 b. P o t h o f f and R o y ' s m o d e l , G = E L a r g e s t r o o t of S L S _ 1 = .678 n e 678 R e f e r . *,-. Q = .40 to Heck t a b l e s , 1 . o / o 1% c r i t i c a l v a l u e = . 3 0 , p < .01 c . R a o ' s model L a r g e s t r o o t of S, S - 1 = . 4 9 0 h e .49 R e f e r ~ .33 t o Heck t a b l e s , 1% c r i t i c a l v a l u e = . 3 3 , p = .01 The p - v a l u e s a re somewhat l a r g e r than i n 2 . , bu t n e v e r t h e l e s s , even the h y p o t h e s i s t h a t the c u r v e s a r e p a r a l l e l must be r e j e c t e d . Discussion As a l r e a d y o b s e r v e d , the growth c u r v e s f o r the sys tem one o y s t e r l i n e s a r e more homogenous than the c u r v e s f o r the sys tem two o y s t e r l i n e s . But even f o r the sys tem one l i n e s d i f f e r e n c e s were f o u n d , so i t won ' t be r e a s o n a b l e t o i g n o r e l i n e s w i t h i n a p a r t i c u l a r s y s t e m - t y p e c o m b i n a t i o n . The c o n s t a n t term i s g e n e r a l l y l a r g e r f o r the sys tem two c u r v e s than f o r the sys tem one c u r v e s w h i c h s u g g e s t s t h a t b a r n a c l e s on s y s t e m two a r e g e n e r a l l y l a r g e r than t h o s e on sys tem one . The sys tem - 81 -one c u r v e s a r e more l i n e a r than the sys tem two c u r v e s . The a v e r a g e s i z e of b a r n a c l e s on i t ems i n sys tem two a p p e a r s to i n c r e a s e f a i r l y q u i c k l y t o s t a r t w i t h and much l e s s r a p i d l y l a t e r on , whereas the a v e r a g e s i z e of b a r n a c l e s on i t e m s i n sys tem one i n c r e a s e s at a more c o n s t a n t r a t e . We cannot c o n c l u d e however , t h a t b a r n a c l e s on system two t e n d t o grow more q u i c k l y at f i r s t than b a r n a c l e s on sys tem one and l e s s q u i c k l y l a t e r on , as the c u r v e s r e p r e s e n t the n a t u r a l p o p u l a t i o n , no t i n d i v i d u a l b a r n a c l e s . I t i s p o s s i b l e t h a t the d i f f e r e n c e does r e f l e c t d i f f e r e n t g rowth c h a r a c t e r i s t i c s f o r the two s y s t e m s , but i f the p r o b a b i l i t y of d y i n g i s r e l a t e d to s i z e , i t may a l s o r e f l e c t the d i f f e r e n t s u r v i v a l p a t t e r n s f o r the two s y s t e m s . Suppose , f o r examp le , t h a t t h e r e i s a t e n d e n c y f o r the b i g g e r b a r n a c l e s t o d i e f i r s t , then a l a r g e number o f d e a t h s i n a p a r t i c u l a r i n t e r v a l would cause the i n c r e a s e i n average s i z e i n t h a t i n t e r v a l t o be s m a l l e r than would be e x p e c t e d f o r the i n c r e a s e i n s i z e of any i n d i v i d u a l . On system one o y s t e r l i n e s b a r n a c l e s d i e o f f v e r y q u i c k l y between t = 5 and t = 8 weeks , but a t a much s l o w e r r a t e a f t e r t h a t , whereas b a r n a c l e s on system two o y s t e r l i n e s d i e o f f a t a more c o n s t a n t r a t e between t = 5 and t = 17 weeks . T h i s c o u l d a c c o u n t f o r the more l i n e a r c u r v e s f o r sys tem one l i n e s . The M o d e l s The e s t i m a t e s of J; a r e v e r y s i m i l a r f o r a l l t h r e e methods . As a l r e a d y p o i n t e d o u t , the c h o i c e o f G i n P o t h o f f and R o y ' s model d o e s n ' t g r e a t l y a f f e c t the e s t i m a t e o f £, but i t does a f f e c t i t s v a r i a n c e and hence the p - v a l u e s of the t e s t s . The p - v a l u e o b t a i n e d u s i n g G = I - 82 -was usually considerably d i f f e r e n t from that obtained using G = E Q . The p-value obtained with G = I tended to agree more cl o s e l y with the A p-value obtained under Rao's model. So the p-value obtained with G = E r o may be suspect, p a r t i c u l a r l y since the estimate of the covariance matrix obtained from the system one oyster l i n e s turned out to be very d i f f e r e n t from that obtained from the system two oyster l i n e s . The estimates of the vector of variances obtained respectively from the system one oyster l i n e s and the system two oyster l i n e s were: [.015, .018, .016, .014, .0096, .026, .035, .051, .040] and [.025, .024, .041, .020, .012, .0094, .018, .014, .024]. The estimates of the c o r r e l a t i o n matrix were respectively: 1 .62 1 .48 .61 1 .44 .57 .55 1 .23 .32 .29 .43 1 .13 .27 .47 .41 .40 1 .20 .25 .40 .37 .23 .77 1 .19 .23 .19 .36 .07 .62 .69 1 .03 .09 .17 .28 .09 .19 .19 .26 and - 83 -1 .72 1 .74 .73 1 .52 .59 .61 1 .54 .61 .58 .68 1 .47 .54 .49 .55 .63 1 .24 .22 .13 .30 .27 .31 1 .35 .35 .35 .50 .57 .61 .47 1 .25 .22 .15 .29 .32 .43 .45 .59 O b s e r v a t i o n s f o r the sys tem one o y s t e r l i n e s a r e more v a r i a b l e towards the end of the s tudy than at the b e g i n n i n g , whereas measurements f o r the sys tem two o y s t e r l i n e s a r e more v a r i a b l e a t t h e b e g i n n i n g o f the s t u d y than l a t e r o n . The two c o r r e l a t i o n m a t r i c e s have a r o u g h l y s i m i l a r p a t t e r n a l t h o u g h the c o r r e l a t i o n s between o b s e r v a t i o n s at t = 7 and t = 8 weeks and between o b s e r v a t i o n s at t = 8 and t = 9 weeks a r e p a r t i c u l a r l y low f o r sys tem two and p a r t i c u l a r l y h i g h f o r sys tem one . In v iew of t h e s e d i f f e r e n c e s i t may be t h a t u s i n g an e s t i m a t e o f £ 0 o b t a i n e d f rom one system as the c h o i c e of G i n the model f o r the o t h e r s y s t e m , l e d to d u b i o u s p - v a l u e s . 3.7 A p p l i c a t i o n of the Model to A l l Seventeen Lines The next s t e p was to i n c o r p o r a t e t h e n e t t i n g and d o w l i n g i n t o the mode l and to t e s t f o r d i f f e r e n c e s not o n l y between the two systems but - 84 -also between the three types of material. Since differences had been found between the system one oyster l i n e s and between the system two oyster l i n e s , each of the seventeen l i n e s was allowed d i f f e r e n t parameters. After f i t t i n g growth curves for each of the seventeen l i n e s , contrasts were examined. This time only Rao's model was used. Since no estimate of E 0 was a v a i l a b l e from a s i m i l a r but independent experiment i n t h i s case, the only reasonable choice of G i n Pothoff and Roy's model would be I, which could d i f f e r s u b s t a n t i a l l y from £ 0 . Y Q (141 x 9) i s the data matrix for a l l 17 l i n e s . The data for system one appears f i r s t , then the data for system two. P i s unchanged. A (141x17) = 5{ 8{ 10f 1 0 0 . . 1 . . 1 . . 0 . 1 0 1 . 0 . 0 . 0 . 0 . 0 . 0 . 0 0 q = . . . . o 1 141 (number of i n d i v i d u a l s ) , m 9 (number of time points), r 17 (number of groups), 17 (rank A). - 85 -At t h i s stage i t was assumed that each set of nine measurements had the same covariance matrix, whether measurements were averages over f i v e as for oyster and dowling, or over ten as for netting. Suppose we wanted an o v e r a l l test for differences between the growth curves for the two systems. This could be done by t e s t i n g the hypothesis: 8Ll + 8 l 2 + ••• + 8 l 9 - 9 I l O " ... ~ 9 J L i 7 = 0 where l± = [?1Q 5±1 . . . l± ^1 i s the set of p parameters for group i . Since the f i r s t nine groups correspond to system one l i n e s , and the other eight to system two l i n e s , t h i s i s equivalent to t e s t i n g that the average of the s over the system one l i n e s i s the same as the average of the C j / s over the system two l i n e s . A r e j e c t i o n of th i s hypothesis would indicate a difference between the two systems i n the average value of at least one of the parameters. The hypothesis can be written C£V = 0 where C = (8 8... 8 -9 -9 ... -9) and V = I (p x p). (1x17) Other contrasts were examined i n order to determine whether differences existed between the three types of material. F i r s t of a l l , i t was determined what degree of polynomial was needed to f i t the growth curves f o r a l l 17 l i n e s adequately. This was done i n the same way as for the oyster l i n e s previously. I t was found that quartics were needed. The f i t t e d growth curves obtained for Dl, D2, D3 were respectively: - 86 -y(t ) = - .0002 tH + .0067 t 3 - .086 t 2 + .578 t + .44 y(t ) = .0005 tk - .0143 t 3 + .126 t 2 - .224 t + 1.06 y ( t ) - .00001 t 4 + .0009 t 3 - .030 t 2 + .387 t + .51 and for Nl and N2: y ( t ) = .0001 tH - .0023 t 3 + .014 t 2 + .100 t + 1.24 y(t) - .0002 t 1* - .0059 t 3 + .058 t 2 - .113 t + 1.62 Pictures of the curves f o r the dowling l i n e s appear i n Figure 3 and for the netting l i n e s i n Figure 4. The curves for the netting l i n e s are almost l i n e a r . A test of the hypothesis of l i n e a r i t y f o r these curves alone yielded a p-value as large as 7%. We have already ascertained that quadratics are adequate for the oyster l i n e s . C l e a r l y i t i s because of the dowling l i n e s , and because of D2 i n p a r t i c u l a r , that I. quartics are needed. The c o e f f i c i e n t of t i s much larger for D2 than for any other l i n e and the growth curve for D2 looks remarkably d i f f e r e n t from any of the other growth curves (see Figure 2). Before discussing t h i s , r e s u l t s of the tests comparing the curves for the two systems and for the three material types w i l l be presented: 1. The growth curves f o r the two systems were compared by t e s t i n g CIV = 0 where C = ( 8 8 8 -9 -9 ... -9) and V = I (5x5), (1x17) Largest root of S,S ~ 1 = .732, h e Refer 16.9 to F 5 1 1 6 , p < .001. - 87 -Figure 3. Growth Curves for the Dowling Lines Figure 4. Growth Curves for the Netting Lines ^ i I 1 1 I 1 1 1 1 1 1 1 1 1 1 0.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 TIME (WEEKS) - 88 -2. The growth curves for the oyster and dowling l i n e s were compared by t e s t i n g C£V = 0 where C = (1 1 1 1 1 1 -4 -4 0 1 1 1 1 1 1 -4 0) and V = I- (5x5). Largest root of S,S - 1 = 1.121-, h e Refer 25.9 to F 5 ) 1 1 6 , p < .001. 3. The growth curves for the oyster and netting l i n e s were compared by te s t i n g C^V = 0 where C = (1 1 1 1 1 1 0 0 - 6 1 1 1 1 1 1 0 - 6 ) and V = I (5x5). Largest root of S.S - 1 = .426, h e ' Refer 9.9 to F 5 > 1 1 6 , p < .001. 4. The growth curves for the dowling and netting l i n e s were compared by t e s t i n g C£V = 0 where C = ( 0 0 0 0 0 0 2 2 - 3 0 0 0 0 0 0 2 - 3 ) and V = I (5x5). Largest root of S h S e _ 1 = .524. Refer 12 to F 5 } n 6 , p < .001. The same analyses were then repeated t h i s time allowing the curves to d i f f e r by an additive constant. So this time £ Q was not included; f o r each test C remained unchanged, but V was replaced by 0 0 0 0 1 0 0 0 0 1 0 0 0 0 1 0 0 0 0 1 - 89 -Results 1. The two systems L a r g e s t r o o t of S, S - 1 = .412 h e R e f e r 12 .2 t o F 4 ) 1 1 9 p < .001 . 2. O y s t e r and d o w l i n g L a r g e s t r o o t of S , S - 1 = . 7 3 0 h e R e f e r 21 .7 t o F 4 > 1 1 9 p < .001 . 3 . O y s t e r and n e t t i n g L a r g e s t r o o t of S - 1 = .35 7 R e f e r 10 .7 t o Yh 1 1 9 p < .001 . 4 . D o w l i n g and n e t t i n g L a r g e s t r o o t of S . S - 1 = .471 h e R e f e r 13 .9 t o Fi+ p < .001 . Discussion D i f f e r e n c e s a r e found b o t h between the two systems and between the t h r e e t y p e s of m a t e r i a l . T h i s i s i n c o n t r a s t to what was found f o r t h e - 90 -s u r v i v a l d i s t r i b u t i o n s , namely that only system was an important f a c t o r . With such a large amount of data i t i s not su r p r i s i n g that d i f f e r e n c e s are found everywhere. The r e l a t i v e size of the test s t a t i s t i c s indicates that the biggest differences l i e between dowling and the other two types of material. This was to be expected i n view of the unusual growth curve for D2. For the comparisons of the two systems and of oyster and dowling l i n e s the p-values were much larger when the curves were allowed to d i f f e r by an additive constant. On the other hand for the other two comparisons the p-values did not change s u b s t a n t i a l l y when the curves were allowed to d i f f e r by a constant. So the difference between netting and dowling and between netting and oyster i s (almost) s o l e l y a t t r i b u t a b l e to the d i f f e r e n t shapes of the growth curves. The difference between the two systems and between oyster and dowling i s due pa r t l y to the d i f f e r e n t shapes of the curves and p a r t l y to d i f f e r e n t l o c a t i o n s . We conclude that barnacles on system two are generally bigger than those on system one and that barnacles on oyster l i n e s are generally bigger than those on dowling l i n e s . The smoothness of the growth curves for the netting l i n e s may r e f l e c t the averaging over ten instead of f i v e , rather than a growth c h a r a c t e r i s t i c . The average size of barnacles on netting l i n e s appears to increase at an almost constant rate except towards the end of the study when i t increases a l i t t l e f a s t e r , whereas the average size of barnacles on dowling l i n e s i s p a r t i c u l a r l y small i n i t i a l l y , increases rapidly to - 91 -s t a r t w i t h and l e s s r a p i d l y l a t e r on . The b e h a v i o r o f the a v e r a g e s i z e of b a r n a c l e s on o y s t e r l i n e s l i e s somewhere between t h e s e two e x t r e m e s . The a v e r a g e s i z e o f b a r n a c l e s on D2 i s e x c e p t i o n a l l y s m a l l a t t = 2 weeks , i n c r e a s e s f a i r l y r a p i d l y u n t i l t = 8 weeks , appears t o r e m a i n the same u n t i l t = 12 weeks and t h e n t o i n c r e a s e e x t r e m e l y r a p i d l y . T h i s c o u l d be a growth c h a r a c t e r i s t i c , but s i n c e t h i s b e h a v i o r i s so d i f f e r e n t , i t may a l s o be t h a t p a r t i c u l a r l y s m a l l b a r n a c l e s were sampled at t = 8 and 9 weeks and p a r t i c u l a r l y l a r g e b a r n a c l e s were sampled at t = 14 weeks . 3.8 Extension of the Model to Allow D i f f e r e n t Covariance Matrices f o r  D i f f e r e n t Items An e s t i m a t e of E Q was o b t a i n e d s e p a r a t e l y f rom the o y s t e r l i n e s , t h e d o w l i n g l i n e s and the n e t t i n g l i n e s . Not s u r p r i s i n g l y the e s t i m a t e o b t a i n e d f rom the n e t t i n g l i n e s was c o n s i d e r a b l y s m a l l e r than the o t h e r e s t i m a t e s s i n c e measurements f rom i t e m s on n e t t i n g l i n e s a r e a v e r a g e s over t e n i n s t e a d of f i v e . I t would seem more r e a s o n a b l e t o assume t h a t t h e c o v a r i a n c e m a t r i x f o r i t e m s on o y s t e r or d o w l i n g l i n e s i s (1/5)E f o r some E, and (1/10)E f o r i t e m s on n e t t i n g l i n e s . The growth c u r v e mode ls can be e x t e n d e d to the case where the c o v a r i a n c e m a t r i c e s f o r d i f f e r e n t i n d i v i d u a l s a r e assumed t o be known m u l t i p l e s of each o t h e r ( I t o , 1 9 6 8 ) . Suppose t h a t t h e r e a r e N i n d i v i d u a l s who a r e assumed to have c o v a r i a n c e m a t r i c e s G l E, e 2 E, e 3 E,..., where the 9^ a r e s c a l a r s and E i s f i x e d but unknown. Suppose the - 92 -i n d i v i d u a l s are divided into groups with nj i n the j t h group. Individuals i n the same group are assumed to have the same covariance matrix, so 0i = 02 = .... = Qn^ and s i m i l a r l y for the other groups. Define a diagonal matrix of weights W (NxN) 1/0! i / e 2 i / e N Instead of choosing £ to minimise (X - A£)' (X - A£) i t i s now chosen to minimise (X - A £ ) f W (X - Ac;) (weighted least squares). So more weight i s given to i n d i v i d u a l s with smaller covariance matrix. It follows that £ = (A^ W A T 1 A{ W X. Tests of C£V = 0 are based on S = ( q I V)« ( C i R C i ) ~ l ( d I V) where R = (A^ W A j ) " 1 + (AJ W A j ) " 1 AJ W X B X' W A j(A{ W ^ ) -1 and S = V'(P S - 1 P * ) _ 1 V. I f W = I ( t h i s means equal weight i s given to a l l observations) we obtain the same expression for R and £ and hence for S^ as previously. The estimate of E i s given by S, where S=-g Sx+g- s2 + 9" S 3 + .... 1 n^+1 ni+n2 +l and S t i s the estimate of the covariance matrix for group t, namely S^  = Y'[I - A,(A» A , ) " 1 A.1] Y where Y i s the data matrix for group t t o l l l l o O a r - 93 -and A, the design matrix for group t, i s a ( n t x 1) vector of l ' s . Individuals i n group 2 have covariance matrix 9 n i + ^ £> those i n group 3 have covariance matrix 0 , ,,£ and so on. If 0i = 0o = ... = 9„ = 1 n]+n 2+l 1 N ( a l l i n d i v i d u a l s have the same covariance matrix) then S w i l l be as for the standard model, namely Y'[I - A.(A! A . ) - 1 A'] Y where Y and A o i l l 1 o o are the data and design matrices for the complete set of data. The test s based on and S e are then as before. Quartics were again f i t t e d and a l l the tests of the previous section were repeated with the adjusted and S e. In our case S = y [covariance matrix for oysters] + •^•[covariance matrix for dowling] + [covariance matrix for netting] and 0^ = 2, i = 1,...,60 and 69,...,131 (items on oyster and dowling l i n e s ) , 0^ = 1, i = 61,...,68 and 132,...,141 (items on netting l i n e s ) . A, Y Q and P are unchanged. Results The same tests as i n the previous section, to compare the growth curves for various pairs of subgroups were carried out: 1. The two systems Largest root of S h S - 1 = .595 Refer 13.9 to F 5 > 1 1 6 , p < .001 - 94 -2. O y s t e r and D o w l i n g L a r g e s t r o o t of S , S _ 1 = .902 h e R e f e r 20 .9 t o F 5 , 1 1 6 , p < .001 3 . O y s t e r and N e t t i n g L a r g e s t r o o t of S, S - 1 = .610 h e R e f e r 14.2 t o F 5 , i 1 6 , p < .001 4 . N e t t i n g and D o w l i n g L a r g e s t r o o t of S, S - 1 = .640 n e R e f e r 14 .8 t o F 5 > 1 1 6 , p < .001 Comments The e s t i m a t e of £ i s u n a f f e c t e d by i n c l u d i n g W i n the a n a l y s i s , s i n c e w i t h i n each group ( l i n e ) the same we igh t i s g i v e n to a l l i t ems ( t h e same c o v a r i a n c e m a t r i x i s assmed f o r i n d i v i d u a l s i n t h e same g r o u p ) . However the p - v a l u e s of the t e s t s a r e a f f e c t e d . In compar i sons i n v o l v i n g n e t t i n g , the p - v a l u e o b t a i n e d i n the o r i g i n a l unwe ighted a n a l y s i s was l a r g e r . T h i s I s because the e s t i m a t e of the c o v a r i a n c e was t o o b i g because i t was o b t a i n e d f rom a l l i t e m s , whether f rom o y s t e r , n e t t i n g or d o w l i n g l i n e s , and t h e s e were a l l assumed to have the same £ Q . F o r c o m p a r i s o n s not i n v o l v i n g n e t t i n g the o r i g i n a l p v a l u e was - 95 -s m a l l e r . T h i s i s because the e s t i m a t e of the c o v a r i a n c e m a t r i x was t o o s m a l l because i t was o b t a i n e d no t o n l y f rom i t e m s on o y s t e r and d o w l i n g l i n e s but a l s o f rom i t e m s on n e t t i n g l i n e s . The p - v a l u e s o b t a i n e d i n the w e i g h t e d a n a l y s i s s h o u l d be e x p e c t e d t o be more r e l i a b l e . The main c o n c l u s i o n s rema in the same however , namely t h a t d i f f e r e n c e s e x i s t b o t h between the two systems and between the t h r e e t y p e s of m a t e r i a l s w i t h the b i g g e s t d i f f e r e n c e b e i n g between the o y s t e r and d o w l i n g l i n e s . 3 .9 Collapse Over Lines Within Each System-Type Combination A l t h o u g h d i f f e r e n c e s had been found between l i n e s of the same t y p e w i t h i n a s y s t e m , ( p a r t i c u l a r l y w i t h i n sys tem t w o ) , a n o t h e r a n a l y s i s was then c a r r i e d out i n wh ich t h e s e d i f f e r e n c e s were i g n o r e d . I g n o r i n g l i n e s we get a t w o - f a c t o r s i t u a t i o n as p r e v i o u s l y d e s c r i b e d . The number of i t e m s i n each c a t e g o r y i s g i v e n be low: Type System 1 System 2 O y s t e r 38 50 N e t t i n g 22 13 D o w l i n g 8 10 The a n a l y s i s was unwe ighted - t h e same c o v a r i a n c e m a t r i x was assumed f o r a l l i n d i v i d u a l s ( i t e m s ) . I t was assumed t h a t the growth c u r v e f o r an i t e m on the j t h type o f l i n e i n the i t h sys tem was - 96 -+ ( ^ o j + ^ l j fc + ••* + V l j t P _ 1 ) So E[Y Q] = A I P, where A (141 x 5) = 38 22 50 13 10 0 0 £(5xp) = HOI H l l • H02 M-12 • HO 3 Hi 3 • £oi ^11 * ^02 S i 2 • and Y Q and P are as before. H - 1 i i p-1 1 V-i 2 p-1 3 'p-1 1 V l 2 Quartics were fitted and the usual hypotheses were tested, namely that the growth curves for various pairs of subgroups were identical: - 97 -1) and V = 1 ( 5 x 5 ) . 0) and V = 1 ( 5 x 5 ) . 3. O y s t e r and N e t t i n g T e s t C I V = 0 where C = (1 0 -1 0 0) and V = 1 ( 5 x 5 ) . L a r g e s t r o o t of S, S - 1 = .874 h e R e f e r 22 .4 t o F 5 > 1 2 8 , p < .001 4 . N e t t i n g and D o w l i n g T e s t C H = 0 where C = (0 1 -1 0 0) and V = 1 ( 5 x 5 ) . L a r g e s t r o o t of S, S - 1 = 5 .022 n e R e f e r 128.5 t o F 5 ) 1 2 8 , P « .001 A g a i n d i f f e r e n c e s a r e found everywhere w i t h the b i g g e s t d i f f e r e n c e s b e i n g between d o w l i n g and each of the o t h e r two t y p e s . The s m a l l e s t d i f f e r e n c e i s between the two s y s t e m s . However the p - v a l u e s f o r the two c o m p a r i s o n s i n v o l v i n g d o w l i n g a r e s u s p i c i o u s l y s m a l l and t h i s method o f c o l l a p s i n g over l i n e s i s s u s p e c t . 1. The two systems T e s t C I V = 0 where C = (0 0 0 1 -L a r g e s t r o o t of S, S - 1 = . 7 5 2 . n e R e f e r 19.2 t o F 5 > 1 2 8 , P < '001 2. O y s t e r and D o w l i n g T e s t C I V = 0 where C = (1 -1 0 0 L a r g e s t r o o t o f S h S e - 1 = 4 .561 R e f e r 116.7 t o F 5 1 2 8 , P « .001 - 98 -3.10 Missing Data A n o t h e r g e n e r a l i s a t i o n of the model c o u l d a l s o have been u s e d , namely the g e n e r a l i s a t i o n s u g g e s t e d by K l e i n b a u m (1973) t o a l l o w m i s s i n g d a t a . In t h i s way the a s s u m p t i o n t h a t o b s e r v a t i o n s a re t a k e n a t the same t imes f o r a l l i n d i v i d u a l s ( I tems) c o u l d have been r e l a x e d . Under t h i s model i t i s assumed t h a t d a t a i s m i s s i n g at random - whether or n o t an o b s e r v a t i o n i s m i s s i n g i s u n r e l a t e d to the v a l u e i t wou ld have t a k e n . K l e i n b a u m ' s g e n e r a l i s e d model p r o c e e d s as f o l l o w s : The n i n d i v i d u a l s a r e d i v i d e d i n t o u d i s j o i n t groups S i , S 2 , . . . , S u such t h a t i n d i v i d u a l s i n the same group Sj have measurements t a k e n a t the same qj t ime p o i n t s . The number of i n d i v i d u a l s i n Sj i s n j . I n d i v i d u a l s i n d i f f e r e n t g roups may no t have measurements t a k e n a t the same t ime p o i n t s but may have measurements t a k e n at the same number o f t i m e p o i n t s (q j = 9 j ' ) * L e t q be the t o t a l number of t ime p o i n t s . Then f o r each group Sj we have a d e s i g n m a t r i x Aj (n j x m) ( s i n c e w i t h i n each group S j , i n d i v i d u a l s a r e s u b d i v i d e d a c c o r d i n g t o w h i c h of the m " t r e a t m e n t " g r o u p s t h e y b e l o n g t o ) . A l s o f o r e a c h group Sj we have an i n c i d e n c e m a t r i x Bj (q x q j ) of O 's and l ' s i n d i c a t i n g the p o s i t i o n s of m i s s i n g o b s e r v a t i o n s f o r i n d i v i d u a l s i n S j • I f ^ j ( n j x q j ) i s the d a t a m a t r i x f o r the i n d i v i d u a l s i n S j and £ Q i s the c o v a r i a n c e m a t r i x f o r a comp le te s e t of q o b s e r v a t i o n s on any i n d i v i d u a l (assumed the same f o r e v e r y i n d i v i d u a l ) , t h e n f o r each j = 1, 2 , . . . , u we h a v e : - 99 -E [ Y . ] = A . I P B. 3 3 J and f o r each row of Y . t h e c o v a r i a n c e m a t r i x i s B . ' S B . , where E and J 3 ° 3 P a r e as f o r t h e growth c u r v e models f o r comp le te d a t a d e s c r i b e d i n S e c t i o n 3 . 3 . As b e f o r e the rows o f Y j a r e assumed to be m u t u a l l y i n d e p e n d e n t and each row i s assumed t o have a m u l t i v a r i a t e norma l d i s t r i b u t i o n . Under t h i s model i t i s not p o s s i b l e to f i n d the maximum l i k e l i h o o d e s t i m a t e s of £ and Z i n c l o s e d f o r m , but K l e i n b a u m f i n d s some b e s t a s y m p t o t i c a l l y normal (BAN) e s t i m a t o r s w h i c h have the same l a r g e sample p r o p e r t i e s as the maximum l i k e l i h o o d e s t i m a t e s . L e t Z* be the column v e c t o r formed by p u t t i n g the columns o f E, u n d e r n e a t h each o t h e r and l e t y j be the v e c t o r formed by p u t t i n g the columns of Y j u n d e r n e a t h e a c h o t h e r . Then a BAN e s t i m a t o r o f ^* ( w h i c h K l e i n b a u m f i n d s by w r i t i n g the model i n the form of a g e n e r a l u n i v a r i a t e w e i g h t e d l e a s t s q u a r e s model) t u r n s out to be: A u A u A £*= [ Z P B . ( B ! Z B . ) - 1 B. P' ® A ! A . ] " Z [P B , ( B ! Z B . ) _ 1 ® A ! ] y . j = 1 3 3 0 3 j 3 3 j J o j 3 3 where M i n d i c a t e s the g e n e r a l i z e d i n v e r s e of the m a t r i x M and Z Q, an u n b i a s e d and c o n s i s t e n t e s t i m a t o r of Z Q, i s f ound as f o l l o w s : The A ( r , s ) e lement a r g o f £ q i s the u s u a l p o o l e d e s t i m a t e u s i n g o n l y t h o s e i n d i v i d u a l s f o r w h i c h measurements were o b t a i n e d at b o t h t r and t g . In o r d e r to t e s t a h y p o t h e s i s of the form H'F,* = 0 , where H (mp x w) i s of f u l l rank w and H'£* i s e s t i m a b l e ( w h i c h means t h a t i t has an u n b i a s e d e s t i m a t o r l i n e a r i n y where y i s the v e c t o r formed by p u t t i n g - 100 -t h e columns of Y u n d e r n e a t h each o t h e r ) , the t e s t s t a t i s t i c i s g i v e n by W n = ( H ' £*)'[H'[ I P B . ( B ' E B . ) _ 1 B, P» ® A ' A , ] - H ] - 1 ( H ' £*) where i s any BAN e s t i m a t o r o f £* and E q i s any c o n s i s t e n t p o s i t i v e d e f i n i t e e s t i m a t o r of £ 0 . I n p a r t i c u l a r t h e e s t i m a t o r s o b t a i n e d by K l e i n b a u m may be u s e d . Under H Q , W n i s a s y m p t o t i c a l l y d i s t r i b u t e d as X 2 . W i s the a n a l o g u e of S S - 1 i n P o t h o f f and R o y ' s model w n h e 3 d e s c r i b e d i n S e c t i o n 3 . 3 . An a l t e r n a t i v e s t a t i s t i c W n * may be used w h i c h i s the ana logue of S, S - 1 i n R a o ' s m o d e l . W * i s o b t a i n e d by h e n J r e p l a c i n g ( A j ' A j ) i n W n by a more c o m p l i c a t e d m a t r i x Qj whose g e n e r a l i z e d i n v e r s e i s a n a l o g o u s t o R o f R a o ' s m o d e l . Application of the Model to our Data T h i s g e n e r a l i s a t i o n might have been a p p r o p r i a t e f o r our d a t a s e t , because some d a t a was n e g l e c t e d i n o r d e r t o s a t i s f y the a s s u m p t i o n t h a t o b s e r v a t i o n s on a l l i n d i v i d u a l s were t a k e n a t the same t i m e s . Namely the d a t a t a k e n a t t = 1, 10 and 11 weeks was i g n o r e d . Under K l e i n b a u m ' s model t h i s d a t a c o u l d have been i n c l u d e d : The i t e m s would have been d i v i d e d i n t o g roups f i r s t a c c o r d i n g to the t i m e s a t w h i c h d a t a was t a k e n on them. F o r example i t e m s f o r w h i c h d a t a was a v a i l a b l e a t 2, 3 , . . . , 9 , 10 , 14, 17 would fo rm one g r o u p , t h o s e f o r w h i c h d a t a was a v a i l a b l e a t 2, 3 , . . . , 9 , 11 , 14, 17 would form a n o t h e r g roup and so o n . W i t h i n each group i t e m s would be d i v i d e d a c c o r d i n g t o l i n e . Thus the A^ 's would be d e t e r m i n e d . Then p o l y n o m i a l s would have - 101 -been f i t t e d as b e f o r e and the p a r a m e t e r s f o r d i f f e r e n t l i n e s compared. However s i n c e the amount of d a t a n e g l e c t e d was so s m a l l compared t o t h e t o t a l amount of d a t a , and s i n c e a p p l i c a t i o n o f the model would have been q u i t e l a b o r i o u s i t was not c o n s i d e r e d w o r t h w h i l e t o p u r s u e t h i s m o d e l . 3.11 Growth Curves f o r Individual Barnacles I n a l l the p r e v i o u s a n a l y s e s r e l a t i n g t o the growth d a t a , a f t e r a v e r a g i n g over the sample of f i v e ( t e n ) , t h e r e was o n l y one o b s e r v a t i o n at each t ime f o r each i t e m . F o r t h i s r e a s o n i t was not p o s s i b l e to f i t a d i f f e r e n t g rowth c u r v e f o r e a c h i t e m . So our c o v a r i a n c e m a t r i x c o n t a i n e d two components of v a r i a b i l i t y - b a r n a c l e - t o - b a r n a c l e v a r i a b i l i t y w i t h i n i t ems and i t e m t o i t e m v a r i a b i l i t y . I t would be n i c e to f i t the a v e r a g e growth cu rve f o r i n d i v i d u a l s on each i t e m and then to compare the growth c u r v e s f o r the d i f f e r e n t i t e m s w i t h i n a l i n e . Whether t h i s i s p o s s i b l e w i t h the l i m i t e d d a t a t h a t i s a v a i l a b l e w i l l be i n v e s t i g a t e d i n t h i s s e c t i o n . C o n s i d e r j u s t t h e f i r s t o y s t e r l i n e . I d e a l l y a d a t a m a t r i x Y^ wou ld be a v a i l a b l e , e a c h row of w h i c h would c o r r e s p o n d t o one b a r n a c l e . T h e r e would be f i v e rows c o r r e s p o n d i n g to e a c h i t e m on the l i n e . I n f a c t a l l t h a t i s a v a i l a b l e i s a d a t a m a t r i x Y 2 . In Y 2 , f i v e rows c o r r e s p o n d t o e a c h i t e m , but w i t h i n each s e t of f i v e rows , s u c c e s s i v e o b s e r v a t i o n s i n a row may o r may not c o r r e s p o n d to the same b a r n a c l e . F u r t h e r m o r e o b s e r v a t i o n s at d i f f e r e n t t ime p o i n t s i n d i f f e r e n t rows may - 102 -correspond to the same barnacle, so the rows are dependent. However, the f i v e observations at any one time are independent. But E[Y]J = E[Y 2] = A£P. So we can use our data matrix Y 2 j u s t as we would have used Y^ to obtain a least squares estimate of £. This estimate involves only the average of the observations at each time and i t doesn't matter whether or not the observations at successive times correspond to the same barnacles - our estimate i s s t i l l v a l i d . Of course i f we were to assume that a l l items have the same growth curve, we would obtain the same growth curve for the oyster l i n e as previously when we f i r s t averaged within each item. But by choosing a d i f f e r e n t design matrix A, we can now obtain a d i f f e r e n t growth curve for each item. Testing for differences between the growth curves i s not as straightforward as f i t t i n g the growth curves: i n order to test hypotheses r e l a t i n g to the parameters we need to f i n d matrices S^ and S~. These involve S and S, an estimate of the covariance matrix S c o for the set of observations on an i n d i v i d u a l barnacle. We have already obtained £, but S i s more d i f f i c u l t . Since each row of Y 2 does ot necessarily correspond to the same barnacle, Y 2 ' [ I - A^CA^' A ^ ) - 1 A]/] Y 2 does not provide an estimate of E Q. However i f a strong assumption i s made about the form of E Q , i t i s possible to obtain an estimate of i t . - 103 -Obtaining an Estimate of E n I f we assume t h a t v a r i a n c e i s c o n s t a n t over t ime and t h a t the c o r r e l a t i o n between o b s e r v a t i o n s d weeks a p a r t on an i n d i v i d u a l i s t h e n E Q i s p r o p o r t i o n a l t o : P 1 P P 1 12 P P 1 P P 1 P P 1 P P 1 P P 1 P 1 Under t h i s s t r o n g a s s u m p t i o n a l l t h a t i s needed i s an e s t i m a t e o f p . T h i s i s o b t a i n e d as f o l l o w s : A number of i t e m s were e x c l u d e d f rom the p r e v i o u s a n a l y s e s because towards the end of the s t u d y , fewer than f i v e b a r n a c l e s were s t i l l a l i v e on them. I t was assumed t h a t t h e s e b a r n a c l e s were r e p r e s e n t a t i v e o f a l l the b a r n a c l e s and an e s t i m a t e of p was o b t a i n e d u s i n g o n l y t h e s e b a r n a c l e s . Suppose t h a t on a p a r t i c u l a r i t e m , f o u r b a r n a c l e s rema ined f o r the l a s t few t ime p o i n t s . Then c l e a r l y we would be s a m p l i n g the same f o u r b a r n a c l e s a t each of t h e s e t i m e s . F u r t h e r m o r e i t was u s u a l l y c l e a r w h i c h o b s e r v a t i o n s at s u c c e s s i v e t i m e s c o r r e s p o n d e d t o t h e same b a r n a c l e s ( s i n c e b a r n a c l e s cannot get s m a l l e r ) . I f t h i s w a s n ' t o b v i o u s - 104 -i t was assumed t h a t the s m a l l e s t o b s e r v a t i o n at the f i r s t t ime c o r r e s p o n d e d to the s m a l l e s t o b s e r v a t i o n at the next t i m e , and so o n . Items w i t h o n l y one b a r n a c l e l e f t c o u l d not be u s e d , as r e p l i c a t i o n was n e e d e d . E l e v e n i t ems ( a l l f rom the sys tem one o y s t e r l i n e s ) were u s e d . L o n g i t u d i n a l d a t a was a v a i l a b l e a t t = 11, 14 and 17 weeks on 2, 3 or 4 b a r n a c l e s on each of t h e s e i t e m s ; the t o t a l number of b a r n a c l e s i n v o l v e d was 34. U s i n g t h i s d a t a , the maximum l i k e l i h o o d e s t i m a t e of p was o b t a i n e d . Koopmans (1942) d e s c r i b e s how t o f i n d the maximum l i k e l i h o o d e s t i m a t e o f p under our model of s e r i a l c o r r e l a t i o n . S i n c e our l o n g i t u d i n a l d a t a was t a k e n at t h r e e week i n t e r v a l s we l e t p' = p . Our a s s u m p t i o n about E Q i m p l i e s t h a t (y t - n t) = P'iy^ - ) + z t where the Z t a r e i n d e p e n d e n t l y d i s t r i b u t e d as N(0, a2), y t and y t _ l a r e o b s e r v a t i o n s on t h e same i n d i v i d u a l r e s p e c t i v e l y at t and t-1, and u t = E [ y t ] , I t f o l l o w s t h a t the v a r i a n c e of the y ' s i s a 2 / ( l - p ' 2 ) . The l i k e l i h o o d of o b t a i n i n g the y ' s i s , (1 " P ' 2 ) N / 2 2 2 L = 2 NT/2 e x P [ " ( A " 2 p ' B + ( 1 + P' )C)/2a ] (2itcf ) where T i s the t o t a l number of t ime p o i n t s , N i s the t o t a l number of i n d i v i d u a l s and A = E £ [ ( y ^ - H^) 2 + ( y * j - i4>2], i j B = E S [ ( y « - uJ) ( y * J - u*) + . . . + ( y j ^ - ^ ) ( y ^ - p £ ) ] , i j C = E E [ ( y ^ j - \L\)2 + . . . ( y j ^ - \ ^ _ ] ) 2 ] , i j - 105 -where y t J i s the o b s e r v a t i o n at t ime t on the j t h b a r n a c l e on the i t h i t e m and p.* = E [ y ^ ] . M a x i m i s i n g t h i s l i k e l i h o o d over the p^ a n d P 1 l e a d s t o an e s t i m a t e o f p ' as the r o o t o f a c u b i c e q u a t i o n (Koopmans, 1 9 4 2 ) . R e s u l t s The e s t i m a t e o f p ' t u r n e d out to be . 8 9 9 . So p = . 9 6 5 . The c o r r e l a t i o n between o b s e r v a t i o n s on the same i n d i v i d u a l a t s u c c e s s i v e t i m e s i s seen t o be v e r y h i g h . I t i s p o s s i b l e t o o b t a i n e s t i m a t e s o f the b a r n a c l e - t o - b a r n a c l e v a r i a n c e w i t h i n i t e m s a t e a c h t ime and thus check our a s s u m p t i o n of c o n s t a n t v a r i a n c e over t i m e . The v a r i a n c e a t t ime t i s e s t i m a t e d by ( y * j - vb2 ? E N - 1 i J where 1 i s the number of i t e m s . The e s t i m a t e s t u r n e d out t o be . 0 5 4 , . 0 6 9 , . 0 5 6 , . 0 5 4 , . 0 4 8 , . 0 4 9 , . 0 6 0 , .043 and .053 s u g g e s t i n g t h a t our a s s u m p t i o n o f c o n s t a n t v a r i a n c e over t i m e was a f a i r l y weak one . Combin ing t h e s e e s t i m a t e s would then l e a d t o an e s t i m a t e of a and thus t o an e s t i m a t e o f Z 0 of the assumed s e r i a l c o r r e l a t i o n f o r m . A l t h o u g h t h i s i s p r o b a b l y the b e s t e s t i m a t e o f S 0 t h a t can r e a l i s t i c a l l y be o b t a i n e d f rom our l i m i t e d d a t a i t i s u n l i k e l y t o be a good e s t i m a t e f o r s e v e r a l r e a s o n s : f i r s t l y a s t r o n g a s s u m p t i o n was made about the form of F,0 and s e c o n d l y the e s t i m a t e was o b t a i n e d u s i n g o n l y a s m a l l number of b a r n a c l e s and u s i n g o b s e r v a t i o n s o n l y a t the l a s t t h r e e t ime p o i n t s . - 106 -Testing Hypotheses of the Form CgV = 0 A H a v i n g o b t a i n e d an e s t i m a t e E q of E q , i s i t p o s s i b l e t o t e s t A A h y p o t h e s e s of the fo rm C£V = 0 by s u b s t i t u t i n g E q f o r S and £ o b t a i n e d A f rom Y 2 f o r £ i n the e x p r e s s i o n s f o r S, and S ? T h i s i s not c l e a r . h e The t e s t s of h y p o t h e s e s w h i c h have been d i s c u s s e d r e l y on the f a c t A t h a t i f n'v-NCu, CE) where c i s c o n s t a n t , and D ~ w p ( k , £)> t n e W i s h a r t d i s t r i b u t i o n ' w i t h k degrees of f r e e d o m , t h e n the d i s t r i b u t i o n o f — 1 2 k 1 u' D u i s known (namely H o t e l l i n g ' s T = — u D _ u). Suppose an e s t i m a t e of £ i s o b t a i n e d f rom the d a t a m a t r i x Y i ( t h e t r u e l o n g i t u d i n a l d a t a ) , namely A t = ( A J A i ) - 1 A J Yl P*(P P ' ) " 1 where we have t a k e n G = I i n P o t h o f f and R o y ' s m o d e l . Then the A c o v a r i a n c e m a t r i x o f £ i s (P P ' ) " 1 P ( A ! A . ) - 1 E P ' ( P P ' ) " 1 = (A! A , ) - 1 E 1 1 o 1 1 where £ i s the c o v a r i a n c e m a t r i x f o r one row of X = Y P ' (P P ' ) - 1 . F u r t h e r A £ ~ N(£, ( A J A j ) " 1 E ) (1) and X ' [ I - A X ( A ^ A j ) - 1 A J ] X ~ W(k, E ) . ( 2 ) T o g e t h e r (1) and (2) i m p l y t h a t the d i s t r i b u t i o n of S^1 i s known and t h i s forms the b a s i s f o r the t e s t s of C£V = 0 . Now suppose t h a t the e s t i m a t e of £ i s o b t a i n e d i n s t e a d f rom our A d a t a m a t r i x Y2 ( l e t t h i s e s t i m a t e be denoted £1) and t h a t - 107 -S = Y ' [ I - A . ( A ! A . ) A ' ] Y i s r e p l a c e d by our e s t i m a t e E o f E . o i l l l o O O Then t h e r e a r e two r e a s o n s why the same d i s t r i b u t i o n t h e o r y no l o n g e r a p p l i e s : A, 1. A l t h o u g h Ci s t i l l has a m u l t i v a r i a t e n o r m a l d i s t r i b u t i o n w i t h mean 5, i t s c o v a r i a n c e m a t r i x i s not the same as t h a t of E, and i n p a r t i c u l a r i s not p r o p o r t i o n a l t o E . 2. E Q does no t have a W i s h a r t d i s t r i b u t i o n as S does because of t h e s p e c i a l form t h a t was assumed f o r i t . So a l t h o u g h i t i s p o s s i b l e to o b t a i n growth c u r v e s f o r i n d i v i d u a l b a r n a c l e s w i t h o u t l o n g i t u d i n a l d a t a , i t does not seem p o s s i b l e to t e s t f o r d i f f e r e n c e s . 3.12 D i s c u s s i o n The growth c u r v e models t h a t were a p p l i e d the p r e v i o u s s e c t i o n s to our l o n g i t u d i n a l d a t a f o r i t ems r e q u i r e d a number of a s s u m p t i o n s . As m e n t i o n e d i n S e c t i o n 3.4 t h e o n l y a s s u m p t i o n t h a t was l i k e l y t o cause a p r o b l e m was the a s s u m p t i o n of e q u a l c o v a r i a n c e m a t r i c e s f o r e v e r y i n d i v i d u a l ( i t e m ) . T h e r e were two r e a s o n s why t h i s might be a s t r o n g a s s u m p t i o n even i f the c o v a r a n c e m a t r i c e s were assumed to be the same f o r e v e r y i n d i v i d u a l b a r n a c l e : 1. Our o b s e r v a t i o n s were a v e r a g e s o v e r f i v e b a r n a c l e s i n some c a s e s and over t e n b a r n a c l e s i n o t h e r c a s e s . 2. At each t ime the p o p u l a t i o n f rom w h i c h the b a r n a c l e s were sampled was of a d i f f e r e n t s i z e f o r each d i f f e r e n t i t e m . - 108 -I n S e c t i o n 3 .8 t h e a n a l y s i s was m o d i f i e d t o a l l o w f o r the f a c t t h a t o b s e r v a t i o n s were a v e r a g e s over sometimes f i v e and sometimes t e n b a r n a c l e s . New p - v a l u e s were o b t a i n e d w h i c h were s l i g h t l y d i f f e r e n t f rom t h o s e o b t a i n e d p r e v i o u s l y . However the b road c o n c l u s i o n s rema ined t h e same. I t i s not so s t r a i g h t f o r w a r d to accommodate the d i f f e r e n t p o p u l a t i o n s i z e s , e s p e c i a l l y s i n c e the number of b a r n a c l e s a l i v e a t each t ime on each i t e m i s no t known (we know o n l y the t o t a l number on e a c h l i n e ) . In S e c t i o n 3.6 e s t i m a t e s of the c o v a r i a n c e m a t r i c e s o b t a i n e d r e s p e c t i v e l y f rom the sys tem one o y s t e r l i n e s and f rom the sys tem two o y s t e r l i n e s "were p r e s e n t e d . The two e s t i m a t e s d i d i n d e e d show somewhat d i f f e r e n t p a t t e r n s w h i c h may be p a r t l y due t o the d i f f e r e n t p o p u l a t i o n s i z e s . I n v iew of t h i s d i f f i c u l t y , the p - v a l u e s o b t a i n e d i n S e c t i o n s 3 . 5 , 3.6 and 3.7 s h o u l d not be c o n s i d e r e d c o m p l e t e l y r e l i a b l e . However t h e p - v a l u e s t h a t we o b t a i n e d were so s m a l l , t h a t even i f the e r r o r i n v o l v e d were q u i t e l a r g e , we would s t i l l r e a c h the same c o n c l u s i o n s . I n S e c t i o n s 3 . 5 , 3.6 and 3 . 7 , growth c u r v e s f o r i t e m s were f i t t e d . I n t h e s e s e c t i o n s each d a t a p o i n t was an a v e r a g e o v e r a number o f b a r n a c l e s a l i v e a t t h a t t i m e . So our g rowth c u r v e s r e p r e s e n t e d the n a t u r a l p o p u l a t i o n . I n S e c t i o n 3.11 we d i s c u s s e d f i t t i n g growth c u r v e s f o r i n d i v i d u a l b a r n a c l e s . But a g a i n our e s t i m a t e of £ was o b t a i n e d f rom d a t a on the b a r n a c l e s w h i c h happened to be a l i v e at any p a r t i c u l a r t i m e . So a g a i n the c u r v e s r e p r e s e n t e d the n a t u r a l p o p u l a t i o n . I d e a l l y we would l i k e to f i t growth c u r v e s t h a t would r e p r e s e n t the g rowth o f a t y p i c a l i n d i v i d u a l b a r n a c l e between t = 2 and 17 weeks . I t - 109 -i s not c l e a r t h a t t h i s i s p o s s i b l e w i t h the l i m i t e d d a t a a v a i l a b l e . I d e a l l y the f o l l o w i n g i n f o r m a t i o n would be a v a i l a b l e f o r e a c h b a r n a c l e : ( a ) Growth d a t a at each t ime p o i n t u n t i l t = 17 weeks or u n t i l i t s d e a t h w h i c h e v e r i s the e a r l i e r . ( b ) Time of d e a t h i f t h i s i s b e f o r e t = 17 weeks or knowledge t h a t i t s u r v i v e d p a s t t = 17. ( c ) Cause of d e a t h - f o o d s h o r t a g e or p r e d a t o r . (d) Number of the i t e m and l i n e t o w h i c h i t i s a t t a c h e d . I f t h i s were the case s e v e r a l a p p r o a c h e s would be p o s s i b l e d e p e n d i n g on the o b j e c t i v e s : 1. Suppose we wanted t o f i t growth c u r v e s f o r i n d i v i d u a l b a r n a c l e s w h i l e a c c o u n t i n g f o r the c e n s o r i n g . The d e a t h s would a f f e c t the a n a l y s i s o n l y i n one way, namely the growth d a t a f o r each i n d i v i d u a l would be t r u n c a t e d a t i t s d e a t h t ime i f t h i s i s b e f o r e t = 17 weeks; we would have growth d a t a w i t h m i s s i n g o b s e r v a t i o n s . K l e i n b a u m ' s model i s s u f f i c i e n t l y g e n e r a l t o a l l o w t h i s p a t t e r n of m i s s i n g d a t a , so t h i s model c o u l d be used to f i t growth c u r v e s and to t e s t f o r d i f f e r e n c e s between v a r i o u s g r o u p s . The growth c u r v e s t h a t we would o b t a i n would r e p r e s e n t not the n a t u r a l p o p u l a t i o n , but the growth e x p e r i e n c e of a t y p i c a l b a r n a c l e between t = 2 and 17 weeks . 2. I f we were i n t e r e s t e d i n s t e a d i n whether s u r v i v a l was r e l a t e d t o g rowth we c o u l d t h i n k of our d a t a as s u r v i v a l d a t a w i t h a t ime dependent c o v a r i a t e . We would have t y p e I c e n s o r i n g i n t h e s u r v i v a l d a t a ; f o r b a r n a c l e s w h i c h s u r v i v e p a s t t = 17 weeks , the e x a c t l i f e t i m e - 110 -would not be known. The time dependent covariate could be s h e l l length. Time dependent covariates can be incorporated into Cox's proportional hazards model, which i s described i n Section 2.1. The hazard function would then be \(t; z ( t ) l = X (t) e 2 ^ ^ where i n our o case: \ Q ( t ) i s an unknown baseline hazard function, B i s an unknown parameter, z(t) Is our time dependent covariate - s h e l l length at time t. In t h i s context a test of 8 = 0 would be a test that growth c h a r a c t e r i s t i c s and s u r v i v a l time are not r e l a t e d . The method of t e s t i n g H Q: 8 = 0 i s described i n Section 2.1. A r e j e c t i o n of t h i s hypothesis would suggest some dependency between growth c h a r a c t e r i s t i c s and s u r v i v a l time, for example a tendency for larger barnacles to l i v e longer. 3 . There are two possible causes of death for the barnacles, namely they may be eaten by predators or they may die due to food shortage. Instead of t e s t i n g whether growth i s related to s u r v i v a l time, i t may be of i n t e r e s t to test separately whether each cause of death i s related to growth. Then we would be able to address questions such as "do the bigger barnacles tend to be eaten f i r s t by the predators?" and " i s i t the smaller barnacles which tend to die of food shortage?" This can be done i n the context of Cox's proportional hazards model by f i n d i n g the hazard function s p e c i f i c to each cause of death. Growth - I l l -i s a g a i n i n c l u d e d as a t ime dependent c o v a r i a t e but t h i s t ime 8 may v a r y o v e r the two t y p e s of f a i l u r e . So e s t i m a t e s of 8^  and P 2 would be o b t a i n e d and a t e s t of B^ = 0 would be a t e s t t h a t the i t h cause of d e a t h i s u n r e l a t e d to growth . To o b t a i n an e s t i m a t e o f 8^, a l l f a i l u r e s o t h e r than t h o s e of t ype i a r e t r e a t e d as c e n s o r e d o b s e r v a t i o n s , and the u s u a l maximum l i k e l i h o o d methods a r e u s e d . U n f o r t u n a t e l y we do not have a l l the i n f o r m a t i o n t h a t we would l i k e : a t each t ime p o i n t we do not have growth d a t a on a l l b a r n a c l e s , o n l y on f i v e f rom e a c h i t e m . F u r t h e r m o r e i n d i v i d u a l b a r n a c l e s cannot be i d e n t i f i e d so i t i s not known whether or n o t s u c c e s s i v e o b s e r v a t i o n s c o r r e s p o n d t o the same b a r n a c l e s , nor w h i c h dea ths c o r r e s p o n d t o w h i c h growth measurements . The cause of d e a t h i s not known. S u r v i v a l d a t a i s a v a i l a b l e o n l y f o r the l i n e s not f o r the i t e m s w i t h i n the l i n e s . F o r t h e s e r e a s o n s we cannot pursue any of the a p p r o a c h e s d e s c r i b e d a b o v e . However , the growth e x p e r i e n c e of a t y p i c a l b a r n a c l e between t = 2 and 17 weeks on a p a r t i c u l a r l i n e w i l l be w e l l r e p r e s e n t e d by the growth c u r v e t h a t was f i t t e d f o r t h a t l i n e i n S e c t i o n 3 . 7 , p r o v i d e d t h a t the p r o b a b i l i t y of a b a r n a c l e d y i n g i s u n r e l a t e d t o i t s s i z e . T h i s can be e x p l a i n e d as f o l l o w s : i f the b a r n a c l e s t h a t d i e between two t ime p o i n t s a r e n e i t h e r p a r t i c u l a r l y b i g or p a r t i c u l a r l y s m a l l , then the change i n a v e r a g e s i z e o f l i v i n g b a r n a c l e s over t h i s t ime i n t e r v a l w i l l be due s o l e l y t o b a r n a c l e s g r o w i n g , not to the c h a n g i n g p o p u l a t i o n . W i t h our l i m i t e d d a t a i t i s not easy to a s c e r t a i n whether the p r o b a b i l i t y of d y i n g i s r e l a t e d t o s i z e : s i n c e s u r v i v a l d a t a i s a v a i l a b l e o n l y f o r the l i n e s , growth d a t a was a v e r a g e d over i t ems w i t h i n - 112 -e a c h l i n e . So f o r e a c h of the 17 l i n e s a v e r a g e s h e l l l e n g t h and the p e r c e n t a g e of d e a t h s were a v a i l a b l e at each t i m e . U s i n g a l l 17 l i n e s a s i g n i f i c a n t c o r r e l a t i o n was found between s i z e a t week 5 and the p e r c e n t a g e d y i n g between weeks 5 and 6 . On t h o s e l i n e s where a v e r a g e s i z e a t week 5 was l a r g e , a s m a l l e r p e r c e n t a g e tended t o d i e between weeks 5 and 6 . However we cannot c o n c l u d e f rom t h i s t h a t w i t h i n a p a r t i c u l a r l i n e the s m a l l e r b a r n a c l e s tend t o d i e f i r s t . The c o r r e l a t i o n may be due to a t h i r d u n d e r l y i n g f a c t o r r e l a t e d t o b o t h s u r v i v a l and g r o w t h . I t may r e f l e c t , f o r example , a sys tem or t r e a t m e n t ( t y p e of m a t e r i a l ) e f f e c t - b a r n a c l e s on system two tend t o s t a r t d y i n g l a t e r and f u r t h e r m o r e they a r e g e n e r a l l y b i g g e r than b a r n a c l e s on sys tem one . To e l i m i n a t e a p o s s i b l e s y s t e m or t r e a t m e n t e f f e c t , we s h o u l d u s e , f o r example , j u s t the sys tem one o y s t e r l i n e s . However t h e r e a r e o n l y s i x of t h e s e so a c o r r e l a t i o n o b t a i n e d j u s t f r o m t h e s e l i n e s w o u l d n ' t be v e r y m e a n i n g f u l . I f s u r v i v a l d a t a were a v a i l a b l e f o r e v e r y i t e m , we would have a v e r a g e s i z e and p e r c e n t a g e d y i n g at each t ime f o r 60 i t e m s f rom system one o y s t e r l i n e s . In t h i s case i t would even be p o s s i b l e to e l i m i n a t e a p o s s i b l e l i n e e f f e c t by u s i n g j u s t the i t ems f rom one l i n e . Then a s i g n i f i c a n t c o r r e l a t i o n at a p a r t i c u l a r t ime between average s i z e and p e r c e n t a g e d y i n g would c e r t a i n l y s u g g e s t dependency between the p r o b a b i l i t y of d y i n g and s i z e . From a b i o l o g i c a l p o i n t of v iew the e x p e r i m e n t e r e x p e c t e d t h a t s i z e and l i f e t i m e would be u n r e l a t e d . W i t h o u t d e t a i l e d s u r v i v a l d a t a we can o n l y assume t h a t t h i s i s t r u e . One o t h e r p o i n t i s o f i n t e r e s t : i n S e c t i o n 2 .5 i t was n o t e d t h a t t h e r e was a s t r o n g c o r r e l a t i o n between the i n i t i a l number on a l i n e and - 113 -the e s t i m a t e s of 8 and a f o r t h a t l i n e . The c o r r e l a t i o n s u g g e s t e d t h a t on l i n e s w i t h fewer b a r n a c l e s i n i t i a l l y ( e . g . sys tem one o y s t e r l i n e s ) , b a r n a c l e s tended to s t a r t d y i n g sooner and a t a f a s t e r r a t e . We now have the a d d i t i o n a l o b s e r v a t i o n t h a t b a r n a c l e s on sys tem one a r e g e n e r a l l y s m a l l e r i n i t i a l l y than t h o s e on system two. We c o u l d s p e c u l a t e f rom t h e s e o b s e r v a t i o n s t h a t sys tem two i s a more i d e a l s i t e f o r the b a r n a c l e s w h i c h i s why more b a r n a c l e s become a t t a c h e d t o sys tem two. But maybe t h e r e i s n ' t room f o r a l l the b a r n a c l e s on sys tem 2 and t h e y have t o compete f o r a p l a c e . The b i g g e r h e a l t h i e r b a r n a c l e s may have a b e t t e r chance of f i n d i n g a p l a c e on sys tem 2, wh ich would e x p l a i n why the b a r n a c l e s on sys tem two are i n i t i a l l y b i g g e r . The b a r n a c l e s on sys tem one tend to s t a r t d y i n g e a r l i e r and a t a f a s t e r r a t e . A l s o b a r n a c l e s on the d o w l i n g l i n e s a r e p a r t i c u l a r l y s m a l l i n i t i a l l y and s t a r t d y i n g e a r l y and at a f a s t r a t e . The f a c t t h a t the p o p u l a t i o n o f b a r n a c l e s on sys tem one and on d o w l i n g l i n e s d i m i n i s h e s q u i c k l y may be due to the f a c t t h a t the b a r n a c l e s he re a r e s m a l l t o s t a r t w i t h or may be a t t r i b u t a b l e t o a p o o r e r env i ronment - i t i s d i f f i c u l t t o say w h i c h . S i m i l a r l y i t i s d i f f i c u l t t o know whether t o a t t r i b u t e d i f f e r e n c e s i n the growth c u r v e s to the d i f f e r e n t e n v i r o n m e n t s or t o the i n i t i a l d i f f e r e n c e s i n the b a r n a c l e s . - 114 -CONCLUSION A d a t a s e t p r o v i d e d by Mr . H. G o l d b e r g r e l a t i n g t o the s u r v i v a l and g r o w t h o f b a r n a c l e s was examined . N o n p a r a m e t r i c t e s t s , namely the l o g r a n k and the W i l c o x o n , i n d i c a t e d t h a t d i f f e r e n c e s e x i s t e d between the s u r v i v a l d i s t r i b u t i o n s even of l i n e s of the same m a t e r i a l t ype w i t h i n the same s y s t e m . The t e s t s s u g g e s t e d t h a t sys tem had more i n f l u e n c e t h a n m a t e r i a l t y p e on the s u r v i v a l d i s t r i b u t i o n . The e x p o n e n t i a l model was f i t t e d f o r the s u r v i v a l d i s t r i b u t i o n s o f each o f the 17 l i n e s and the e s t i m a t e d s l o p e s , 8 were compared . A random e f f e c t s model was d e v e l o p e d f o r the 8 v a l u e s , i n i t i a l l y i n c o r p o r a t i n g o n l y the o y s t e r l i n e s but s u b s e q u e n t l y i n c l u d i n g a l l 17 l i n e s . Maximum l i k e l i h o o d e s t i m a t e s were o b t a i n e d o f the t r u e u n d e r l y i n g 8 v a l u e f o r l i n e s of each t y p e w i t h i n each s y s t e m . A c o m p a r i s o n of t h e s e e s t i m a t e s s u g g e s t e d t h a t s l o p e was dependent on sys tem but not on m a t e r i a l t y p e . A g a i n sys tem was found t o be the more i m p o r t a n t f a c t o r . The growth d a t a was l e s s s t r a i g h t f o r w a r d to examine as we d i d no t have l o n g i t u d i n a l d a t a . We a v e r a g e d over the s e t of measurements o b t a i n e d at each t ime f rom each i t e m and t r e a t e d the r e s u l t i n g d a t a s e t as l o n g i t u d i n a l d a t a f o r the i t e m s . U s i n g t h i s s e t o f a v e r a g e s , p o l y n o m i a l growth c u r v e s f o r i t ems were f i t t e d t o each l i n e u s i n g b o t h P o t h o f f and R o y ' s model and R a o ' s m o d e l . Q u a d r a t i c s were adequate e x c e p t i n the case of the second d o w l i n g l i n e . A c o m p a r i s o n of t h e p a r a m e t e r s i n d i c a t e d t h a t d i f f e r e n c e s e x i s t e d b o t h between systems and - 115 -between m a t e r i a l t y p e s , w i t h the b i g g e s t d i f f e r e n c e s b e i n g between d o w l i n g and the o t h e r two t y p e s of m a t e r i a l . The p - v a l u e s may be somewhat u n r e l i a b l e as the s t r o n g a s s u m p t i o n of e q u a l c o v a r i a n c e m a t r i c e s f o r e v e r y i n d i v i d u a l ( i t e m ) had to be made. A l t h o u g h , we d i d n ' t have l o n g i t u d i n a l d a t a f o r b a r n a c l e s i t was found t o be p o s s i b l e t o f i t the a v e r a g e growth c u r v e f o r i n d i v i d u a l b a r n a c l e s on each i t e m but not t o t e s t f o r d i f f e r e n c e s between the c u r v e s o b t a i n e d . I n c o n c l u s i o n , a l t h o u g h we had a l a r g e amount of growth d a t a , a s m a l l e r amount of l o n g i t u d i n a l d a t a would have been more u s e f u l . - 116 -BIBLIOGRAPHY B r e s l o w , N . E . ( 1 9 7 4 ) . C o v a r i a n c e a n a l y s i s o f c e n s o r e d s u r v i v a l d a t a . B i o m e t r i c s 30 , 8 9 - 9 9 . Cox , D.R. ( 1 9 7 2 ) . R e g r e s s i o n models and l i f e t a b l e s ( w i t h d i s c u s s i o n ) . J o u r n a l o f the R o y a l S t a t i s t i c a l S o c i e t y B, 34 , 187-202 . G r i z z l e , J . and A l l e n , D.M. ( 1 9 6 9 ) . A n a l y s i s of growth and dose r e s p o n s e c u r v e s . B i o m e t r i c s 25 , 3 5 7 - 3 8 1 . I t o , K. ( 1 9 6 8 ) . On the e f f e c t o f h e t e r o s c e d a s t i c i t y and n o n n o r m a l i t y upon some m u l t i v a r i a t e t e s t p r o c e d u r e s . I n : K r i s h n a i a h , P .R. e d . ; M u l t i v a r i a t e A n a l y s i s I I . New Y o r k : Academic P r e s s , 8 7 - 1 2 0 . K a l b f l e i s c h , J . G . and P r e n t i c e , R . L . ( 1 9 8 0 ) . The s t a t i s t i c a l a n a l y s i s of f a i l u r e t ime d a t a . New Y o r k : W i l e y . K h a t r i , C . G . ( 1 9 6 6 ) . A n o t e on a manova model a p p l i e d t o p rob lems i n g rowth c u r v e . A n n a l s of the I n s t i t u t e of S t a t i s t i c a l M a t h e m a t i c s 18, 7 5 - 8 6 . K l e i n b a u m , D .G . ( 1 9 7 3 ) . A g e n e r a l i z a t i o n of the growth c u r v e model w h i c h a l l o w s m i s s i n g d a t a . J o u r n a l of M u l t i v a r i a t e A n a l y s i s 3, 117-124 . Koopmans, T . ( 1 9 4 2 ) . S e r i a l c o r r e l a t i o n and q u a d r a t i c forms i n n o r m a l v a r i a b l e s . A n n a l s of M a t h e m a t i c a l S t a t i s t i c s 13, 1 4 - 3 3 . L a i r d , N.M. and Ware, J . H . ( 1 9 8 2 ) . Random e f f e c t s models f o r l o n g i t u d i n a l d a t a . B i o m e t r i c s 38 , 963-974 . L a w l e s s , J . F . ( 1 9 8 2 ) . S t a t i s t i c a l models f o r l i f e t i m e d a t a . New Y o r k : W i l e y . L e e , Y . K . ( 1 9 7 4 ) . A n o t e on R a o ' s r e d u c t i o n of P o t h o f f and R o y ' s g e n e r a l i z e d l i n e a r m o d e l . B i o m e t r i k a 6 1 , 3 4 9 - 3 5 1 . Lehmann, E . L . ( 1 9 1 7 ) . N o n p a r a m e t r i c s : s t a t i s t i c a l methods based on r a n k s . H o l d e n Day , San F r a n c i s c o . Machen , D. ( 1 9 7 5 ) . On a d e s i g n p r o b l e m i n growth s t u d i e s . B i o m e t r i k a 3 1 , 7 4 9 - 7 5 3 . M o r r i s o n , D . F . ( 1 9 7 6 ) . M u l t i v a r i a t e s t a t i s t i c a l m e t h o d s . New Y o r k : M c G r a w - H i l l . P e t o , R. ( 1 9 7 2 ) . C o n t r i b u t i o n t o the d i s c u s s i o n o f p a p e r by D.R. Cox. J o u r n a l of the R o y a l S t a t i s t i c a l S o c i e t y B, 34 , 2 0 5 - 2 0 7 . I - 117 -P o t h o f f , R . F . and Roy , S . N . ( 1 9 6 4 ) . A g e n e r a l i z e d m u l t i v a r i a t e a n a l y s i s of v a r i a n c e model u s e f u l e s p e c i a l l y f o r g rowth c u r v e p r o b l e m s . B i o m e t r i k a 5 1 , 3 1 3 - 3 2 6 . P r e n t i c e , R . L . ( 1 9 7 8 ) . L i n e a r rank t e s t s w i t h r i g h t - c e n s o r e d d a t a . B i o m e t r i k a 6 5 , 167-179 . P r e n t i c e , R . L . and M a r e k , P. ( 1 9 7 9 ) . A q u a l i t a t i v e d i s c r e p a n c y between c e n s o r e d d a t a rank t e s t s . B i o m e t r i c s 35 , 8 6 1 - 8 8 6 . R a o , C R . ( 1 9 6 5 ) . The t h e o r y of l e a s t s q u a r e s when the p a r a m e t e r s a r e s t o c h a s t i c and i t s a p p l i c a t i o n to the a n a l y s i s of growth c u r v e s . B i o m e t r i k a 52 , 447-458 . Rao , C R . ( 1 9 6 6 ) . C o v a r i a n c e a d j u s t m e n t and r e l a t e d prob lems i n m u l t i v a r i a t e a n a l y s i s . I n : K r i s h n a i a h , P .R . e d . ; M u l t i v a r i a t e A n a l y s i s . New Y o r k : Academic P r e s s , 8 7 - 1 0 3 . S a t t e r t h w a i t e , F . E . ( 1 9 4 6 ) . An a p p r o x i m a t e d i s t r i b u t i o n of e s t i m a t e s o f v a r i a n c e components . B i o m e t r i c s B u l l e t i n 2, 110-114. Timm, N.H. ( 1 9 8 0 ) . M u l t i v a r i a t e a n a l y s i s o f r e p e a t e d measurements . I n : K r i s h n a i a h , P .R . e d . ; A n a l y s i s of V a r i a n c e . New Y o r k : N o r t h H o l l a n d , 4 1 - 8 7 . 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0096064/manifest

Comment

Related Items