Development and cross-validation of a short form for the WISC-R.

Abstract

The multiple linear regression method was used to derive a short form for the Wechsler Intelligence Scale for Children-Revised in a sample of 300 children receiving treatment and/or evaluation at a university-affiliated community mental health center. The predictive validity of this approach was examined via a crossvalidation that included 300 additional subjects taken from the same population, as well as two short-form regression equations taken from the work of Kennedy and Elder (1982), and a short form derived from the Wechsler normative sample (Kaufman, 1976). Correlations and standard errors of the estimates are reported for both the original and cross-validational samples, and it is pointed out that crossvalidations of this kind result in empirical assessments of shrinkage, as opposed to the theoretical estimates provided by various shrinkage formulae. It is also suggested that there are two distinct types of shrinkage that need to be accounted for in the conduct of short-form research and that, in concentrating on the issue of part-whole shrinkage, prior studies in the area have virtually ignored the issue of cross-validational shrinkage.

Equal weighting vs. differential weighting of subtest scores on short forms of Wechsler's intelligence scales

“…As noted previously, Beck et al (1983) found relatively little shrinkage in the validities of short forms of the WISC-R that involved differential weighting when the regression equation based on data from one sample was applied to data from a different sample. As noted previously, Beck et al (1983) found relatively little shrinkage in the validities of short forms of the WISC-R that involved differential weighting when the regression equation based on data from one sample was applied to data from a different sample.…”

“…Quite apart from my earlier study of short forms of the WISC (Silverstein, 1967), previous studies of short forms of the WISC-R provided a broad hint as to what the findings of the present study might be. Beck et al (1983) reported validities that ranged from .95 to .97 for several short forms, and validities in this same range when their multiple regression equations were cross-validated with data from different samples. The correlation using the two sets of weights was .997.…”

“…The short forms selected for investigation were as follows: -The five-subtest short form of the WISC-R (Information, Block Design, Comprehension, Picture Arrangement, and Coding) first proposed by Kennedy and Elder (1982) and since employed by Beck et al (1983), Phelps and Rosso (1986), Vollmerhausen et al (1986), Wade, Phelps, and Falasco (1986), and Zimet, Farley, and Dahlem (1985). This short form was chosen because of its current popularity and also because Zimet et al (1985) took the unusual step of obtaining multiple regression weights based on their own data.…”

“…-One of the five-subtest short forms of the WISC-R (Vocabulary, Object Assembly, Picture Arrangement, Arithmetic, and Picture Completion) proposed by Beck et al (1983). This particular short form was chosen because it contains only one subtest, Picture Arrangement, in common with Kennedy and Elder's short form.…”

“…Since then, the Wechsler Intelligence Scale for Children-Revised (WISC-R; Wechsler, 1974) has replaced the WISC, and several investigators have proposed short forms of that scale that involve differential weighting (Beck, Ray, Seidenberg, Young, & Gamache, 1983;Clarizio & Veres, 1984;Kennedy & Elder, 1982;Vollmerhausen, Elder, & Clark, 1986). Today, we also have the Wechsler Preschool and Primary Scale of Intelligence (WPPSI; Wechsler, 1967) and the Wechsler Adult Intelligence Scale-Revised (WAIS-R; Wechsler, 1981), and although no one appears to have suggested a short form of either the WPPSI or the WAIS-R that involves differential weighting, the question posed earlier is potentially relevant to short forms of those scales as well.…”