Some notes and speculations about various topics. ||| Gegenwärtig - vorübergehend -
wohl eher eine Gedankensammlung als ein Naturwissenschaftsblog. Das Konzept eines "Naturwissenschaftsblogs" wird erst im kommenden Jahr (2018/2019) Umsetzung finden.
Dienstag, 11. Juli 2017
Brain volume and intelligence: The moderating role of intelligence measurement quality
Correlation between brain volume and IQ in healthy adults is r ≈ .40.
The importance of correcting correlations for range restriction is demonstrated.
Intelligence measurement quality was a moderator of the brain volume/IQ effect.
Fair, good, and excellent measures of IQ yielded correlations of .23, .32, and .39.
p-Curve analysis indicated the significant results in the area likely not due to p-hacking.
A substantial amount of empirical research has estimated the association between brain volume and intelligence. The most recent meta-analysis (Pietschnig, Penke, Wicherts, Zeiler, & Voracek, 2015) reported a correlation of .24 between brain volume and intelligence – notably lower than previous meta-analytic estimates. This headline meta-analytic result was based on a mixture of samples (healthy and clinical) and sample correlations not corrected for range restriction. Additionally, the role of IQ assessment quality was not considered. Finally, evidential value of the literature was not formally evaluated. Based on the results of our meta-analysis of the Pietschnig et al.'s sample data, the corrected correlation between brain volume and intelligence in healthy adult samples was r = .31 (k = 32; N = 1758). Furthermore, the quality of intelligence measurement was found to moderate the effect between brain volume and intelligence (b = .08, p = .028). Investigations that used ‘fair’, ‘good’, and ‘excellent’ measures of intelligence yielded corrected brain volume and intelligence correlations of .23 (k = 9; N = 547), .32 (k = 10; N = 646), and .39 (k = 13; N = 565), respectively. The Henmi/Copas adjusted confidence intervals, the p-uniform results, and the p-curve results failed to suggest evidence of publication bias and/or p-hacking. The results were interpreted to suggest that the association between in vivo brain volume and intelligence is arguably best characterised as r ≈ .40. Researchers are encouraged to consider intelligence measurement quality in future meta-analyses, based on the guidelines provided in this investigation.