UBC Theses and Dissertations
Multivariate analytic investigations for the identification of sub-populations within aphasia Clark, Campbell McGillivrary
Using medically diagnosed aphasics, this study attempted to develop a classification system, based on statistical or, more precisely, variance criteria. Aphasia is a language disorder resulting from insult to the brain. Besides having direct ramifications on aphasia theory and research strategies, the techniques used here may be applicable to other atypical populations. Specifically, the outlined multivariate procedures can provide insight into intra-population distributions. The resulting distribution can then be used as a basis for factorial designs in future research. The sample consisted of seventy-two medically diagnosed aphasics who had received speech therapy for varying time periods. Prior to and at the termination of treatment, subjects were assessed on a well-standardized and reliable aphasia battery, the Porch Index of Communication Ability. This instrument consists of eighteen subtests and purports to measure three dimensions of communication ability: one verbal, one gestural, and one graphic. Post-treatment subtest scores were factor analyzed to determine the factor structure of the P.I.C.A. and this structure was compared with Porch's formulations. Two subtests were deleted from this analysis due to ceiling effects. Using the factor score coefficients from this analysis, factor scores for the pre-treatment subtests were generated. These factor scores were submitted to a hierarchical cluster analysis in order to determine the optimal number of groups within this sample. A series of step-wise, discriminant analyses, using the factor scores as predictors, confirmed that a six-group solution was best. Group distributions were then examined with respect to current models of aphasia. The findings support Porch's formulations with respect to the underlying demensions of the P.I.C.A. However, the results suggested that 1) two subtests were too easy for all subjects and therefore should be discarded, and 2) two other subtests may not, in fact, reflect symbolic language function. In addition, evidence of a general language factor was also found. These findings were also compared to a factor analysis on Porch's standardization sample. The five derived dimensions of communication ability were : 1) verbal fluency, 2) writing (agraphia), 3) gestural demonstration with, varying input modalities, 4) pantomime, and 5) copying. These dimensions were discussed within the context of previous aphasia research. The grouping analysis indicated that the groups were distributed on a severity continuum not a salient features model of aphasia. Only one of the six groups was suggestive of differential impairment with respect to communication modalities, but this group was small (n = 4) and therefore must be considered judiciously, In addition, the applicability of these procedures was discussed within the context of multivariate research. The power of multivariate designs rests on minimizing the number of groups and dependent variables and maximizing the number of subjects, and the present procedures should aid in meeting these criteria. Specifically, factor analysis reduces the number of dependent variables and yet allows for a representative sample of dependent variables. Cluster analysis groups Individuals Based on the similarities of their performance and thus relatively homogenous groups are formed. The distribution of these groups can form the basis for subsequent factorial designs.
Item Citations and Data