Building a Better IQ: g Loadings of IQs Experimentally Controlled for Subtest Number, Heterogeneity, g Loading Saturation, and Weighting

Intelligence tests produce IQs which are interpreted as indexes of psychometric g. It is thus important to promote the accuracy of those IQs, and to isolate the characteristics of IQs that result in more and less accuracy. In this study, we identified a number of characteristics of IQs that vary across tests, including subtest number, subtest selection methods (e.g., selecting based on heterogeneity, subtest g loading, and combined approaches), and weighting procedures. We created experimentally manipulated IQs to systematically test the influence of these characteristics on IQ g loadings using three archival datasets. Using cross-battery confirmatory analysis, g loadings for each experimental IQ were calculated. Results indicate that increasing subtest number and selecting subtests based on their g loading produce IQs with the highest g loadings while other methods such as heterogeneous subtest selection and weighting had a nonsignificant influence on IQ g loading. The final model accounted for approximately 46% of the variance in IQs that is attributable to psychometric g. Discussion regarding directions for future research and implications for test selection and development is provided