UBC Theses and Dissertations
Studies in applied statistics : ship hull design optimization and endogamous group comparisons Thorne, Anona Elaine
This paper contains two case studies in applied statistics. The first study originated from a series of seakeeping experiments on twenty-seven different ship hull designs. The experimenters sought a method of predicting performance for hull designs different from those included in the study. The results of these experiments had already been analyzed, but we wished to consider the Taguchi method as an alternative method of analysis. In this study, we describe the initial design problem and analyses and provide a critical evaluation of the Taguchi method; we assess its merits to determine its suitability for the ship hull design problem. We find the Taguchi method unsatisfactory in general, and inappropriate for the ship hull design problem. The second study derives from a case which came to the Statistical Consulting and Research Lab (SCARL) in the Department of Statistics at The University of British Columbia. Having lost all her original data, our client wished to know if it were possible to "reconstruct" her data or, in any event, to perform a statistical analysis, using the tables of summary statistics in her possession. She wished to model several different pulmonary functions using various demographic and anthropometric measurements as explanatory variables. For each of the different model variants under consideration, she wanted to see if the same model parameters would be valid for all endogamous groups, or if some groups were sufficiently different that the parameter values would be different for these groups. We were able to develop a new general method for application to her problem; the derivation and application of this methodology are presented in the second study.
Item Citations and Data