Age and Sex Invariance of the International Cognitive Ability Resource

Background. Condon and Revelle (2014) developed the International Cognitive Ability Resource (ICAR) to provide flexible, unrestricted test items to researchers, and published initial validity evidence for the four pilot item... [ view full abstract ]

Background. Condon and Revelle (2014) developed the International Cognitive Ability Resource (ICAR) to provide flexible, unrestricted test items to researchers, and published initial validity evidence for the four pilot item types: letter-number series, matrix reasoning, three-dimensional rotation, and verbal reasoning. In order to meaningfully interpret ICAR scores across salient dimensions of populations such as sex and age, measurement invariance must be established. The present study aims to examine the following research questions: 1) Are the abilities measured by the ICAR 60-item (ICAR60) and 16-item (ICAR16) tests invariant across sex and age groups? And, 2) what is the influence of age and sex on the abilities measured by the ICAR?

Method. To address the first research question, a multiple group confirmatory factor analysis was conducted on the first-order models proposed by Condon and Revelle (2014) using the ICAR norming sample retrieved from the Journal of Open Psychology Data. Participants were a sample of 96,958 individuals (66% female) ages 14 to 90 years (M=26, SD=10.6). Models were tested across the sex and age groups to establish configural, metric, and scalar invariance of the 16- and 60-item tests. Then, a multiple indicator multiple cause (MIMIC) model was fitted to the data to examine the influence of age and sex on the latent factors. The age variable was mean centered and scaled in order to include exponential age variables in the model. 

Results. The ICAR16 and ICAR60 structural models fit the data well for males and females and across age groups based on the fit indices CFI, TLI, and RMSEA (CFI= .966 to .995, TLI =.965 to .994, RMSEA= .005 to .006). Both the forms demonstrated acceptable decreases in CFI (<.01) for loading- and threshold-constrained models across sex and age groupings. The MIMIC model revealed a small but significant difference favoring males on each of the four latent variables across forms (b_female= .130 to .506, p<.001). Age, age squared, and age cubed significantly influenced the letter-number series, matrix reasoning, three-dimensional rotation latent variables on both forms (b_age= .007 to .0215, b_agesq= -.161 to -.044., b_agecube= .011 to .027,  p<.001). 

Discussion. Findings suggest that the ICAR16 and ICAR60 measure the same constructs across sex and age groups, and the same strength of the relations exists among the first-order factors and the items. The MIMIC models revealed a small sex difference across latent factors for both forms. Age had a curvilinear effect on each of the subtests across forms, which is consistent with the current understanding of the development of fluid and spatial reasoning abilities throughout the lifespan (Salthouse, 2009). The direction of the coefficients indicates multiple curves in the data, but this is likely due to an overpowered sample. 

with the exception of verbal ablilities (Schaie, 2017).