A COMPARISON OF THE WISC-R FACTOR STRUCTURES OBTAINED FROM PEARSON VS. STANDARDIZED SUMS AND DIFFERENCES CORRELATIONS
The purpose of the study was to investigate the differences in the WISC-R factor structure that result from using classical statistical methods versus robust statistical methods. The WISC-R protocols for 259 special education children in a low socioeconomic urban area were obtained for the study. The children had been classified as Educable Mentally Retarded, Neurologically Impaired, Emotionally Disturbed and Other. The "other" category included combinations of the three special conditions and Communication Handicapped. Each of the ten standard WISC-R subtest raw scores were adjusted for age and the residuals of that adjustment were then examined for violations to the assumption of normality. Outliers were identified in the distributions of all of the subtests. Having identified violations to the underlying assumptions of classical statistical methods it was then possible to assess (1) the effects on the WISC-R factor structure of these violations and (2) the ability of robust statistical methods to uncover the WISC-R factor structure in the face of non-normal distributions. Four correlation matrices were obtained from the age adjusted subtest scores. The first matrix contained Pearson product moment correlations (classical method) and the remaining three matrices contained Standardized Sums and Differences correlations (robust method). The robust correlation matrices differed in the percent of outliers trimmed from the subtest distributions when calculating the robust correlation. Each of the four correlation matrices were submitted to a principal components factor analysis with a varimax rotation. Two and three factor solutions were retained. The results of the factor analyses for the classical method and the robust method were compared. Implications for the construct validity of the WISC-R were noted. The results indicated that robust methods obtained greater "simple structure" for the three factor solution than the classical method. The greatest impact occurred on Factor III, Freedom from Distractibility, where only Coding and Block Design had significant loadings. The results indicated the appropriateness of the use of the Verbal-Performance format for the sample of urban, low socioeconomic status, special education students examined in this study and also demonstrated that the WISC-R possesses construct validity for this special population.
SQUIRES, PAUL, "A COMPARISON OF THE WISC-R FACTOR STRUCTURES OBTAINED FROM PEARSON VS. STANDARDIZED SUMS AND DIFFERENCES CORRELATIONS" (1982). ETD Collection for Fordham University. AAI8213622.