Validity and Fairness Studies of the Kaufman Adolescent and Adult Intelligence Test: Differential Item Functioning, Factor Structure, and Profile Analysis
Abstract
Three validity and fairness studies of the Kaufman Adolescent and Adult Intelligence Test (KAIT; Kaufman & Kaufman, 1993) were conducted, based on data obtained from the standardization sample (N = 2,000). The instrument's theoretical structure is based on Horn and Cattell's (1966) fluid and crystallized theory of intelligence. The Standard Battery includes Fluid (Rebus Learning, Logical Steps, Mystery Codes) and Crystallized ( Definitions, Auditory Comprehension, Double Meaning) subtests; the Extended Battery includes an additional Fluid (Memory for Block Designs) and Crystallized (Famous Faces) subtest. Composite scores (M = 100, SD = 15) include: Crystallized, Fluid, and Composite. Study One derived and described a normative typology of KAIT core profiles. Results indicated a three-cluster solution meet all statistical criteria (e.g., between-cluster dissimilarity, internal homogeneity, and replicability) for the Standard and Extended Batteries. Core profiles were labeled: Below Average, Average , and Above Average, and did not differ in terms of gender. Study Two used the item response theory likelihood ratio approach, based on a 2-parameter logistic model, to investigate gender differential item functioning (DIF) in KAIT items. Subtest items examined for DIF included: Definitions, Auditory Comprehension, Double Meaning, Famous Faces, Mystery Codes, and Memory for Block Designs. Results indicated the presence of uniform (i,e., difficulty differences) and nonuniform (i.e., discrimination differences) DIF in several of the items in the Crystallized subtests. Study Three used multisample confirmatory factor analysis (MCFA) and latent means structures analysis (LMS) to test factorial invariance and latent mean differences across gender groups. MCFA found that the error variance of Famous Faces lacked invariance. LMS found that the intercepts of Memory for Block Designs, Famous Faces, Auditory Comprehension , and Logical Steps lacked invariance, with effect size estimates indicating that males obtained slightly higher expected scores than females. Consequently, latent mean differences were not conducted. As such, the KAIT demonstrated partial measurement invariance (Byrne, Shavelson, & Múthen, 1989). Collectively, the results of these studies indicate the degree to which KAIT scores can be used as comparable measures of intelligence across gender groups. Implications for test users are provided.
Degree
Ph.D.
Advisors
Maller, Purdue University.
Subject Area
Quantitative psychology
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.