Modern intelligence tests purport to measure the same cognitive constructs, but the subtests within each test battery vary according to task demands, stimuli, and response format. Due to these subtest specific differences, it... [ view full abstract ]
Modern intelligence tests purport to measure the same cognitive constructs, but the subtests within each test battery vary according to task demands, stimuli, and response format. Due to these subtest specific differences, it is questionable whether or not these different tests are actually measuring the same abilities and whether results are comparable across tests. A comprehensive simultaneous analysis of several intelligence tests can address questions concerning the structure of cognitive abilities at a content, rather than test specific, level.
Six intelligence tests (KABC-II, WJ III, WISC-III, WISC-IV, WISC-V, and DAS-II) were included in a cross-battery confirmatory factor analysis (CB-CFA). Participants were 3,930 children and adolescents aged 6 to 16 drawn from seven USA nationally representative standardization linking samples. The sample sizes ranged from 181 to 2,520 per standardization sample. In order to simultaneously analyze several tests a planned missingness methodological approach was used, which addresses concerns regarding bias and power (Enders, 2010; Graham et al., 2006). Samples were linked to each other by tests they share in common.
Analyses were completed in a series of steps. Eight measurement invariance models were tested across different samples and editions of the same tests; all were supported. CB-CFA models included six Cattell-Horn-Carroll (CHC) cognitive broad abilities (Gc, Gf, Gv, Gsm, Gs, and Glr) and g latent variables; each broad ability latent variable was estimated by 7 – 15 subtests.
Preliminary results indicated a cross-battery CHC model fit the data well, suggesting the six CHC broad abilities and g were invariant across the six intelligence tests analyzed in this study. Alternative models, which include a first-order model and models with various cross-loadings, will be compared and discussed.
