Schooling showed no significant interaction with cogPGS or SES on any intelligence domain, and Bayesian ROPE analysis found interaction effects under 0.05 SD with 95% probability¶

Claim · All claims
Evidence · 1 study (2 entries) · causal r2 · q3 · i0 negligible
Currency · newest study 2022 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q3 i0 No significant two-way (schooling-cogPGS, schooling-SES, cogPGS-SES) or three-way interactions were found for cIQ, fIQ, or WM; almost all interaction terms fell entirely within a 0.05 SD ROPE boundary. → Nicholas Judd 2022 q3 i? A 0.02 SD effect could not be ruled out: most interaction distributions substantially overlap the 0.02 ROPE boundary. → Nicholas Judd 2022 (2)

Evidence¶

Nicholas Judd 2022¶

Nicholas Judd, Bruno Sauce and Torkel Klingberg. (2022). Schooling substantially improves intelligence, but neither lessens nor widens the impacts of socioeconomics and genetics. npj Science of Learning. https://doi.org/10.1038/s41539-022-00148-5

q3 · i0 · causal · r2

Post hoc Bayesian hierarchical mixed-effects models of the two-way interaction terms of interest, with weakly informative zero-centered priors. No frequentist interactions were significant for any domain (Eqs. 3 and 4).

"Almost all interaction terms were entirely within a ROPE boundary of 0.05 (Fig. 3). This means that with a 95% probability, the interaction effect was less than 0.05 SD, adding evidence for the null hypothesis"

Nicholas Judd 2022 (2)¶

Nicholas Judd, Bruno Sauce and Torkel Klingberg. (2022). Schooling substantially improves intelligence, but neither lessens nor widens the impacts of socioeconomics and genetics. npj Science of Learning. https://doi.org/10.1038/s41539-022-00148-5

q3 · i? · causal · r2

Bayesian ROPE analysis used two boundaries (0.05 and 0.02) because of a lack of consensus on minimal effect size of interest and because schooling is a cumulative process. The exceptions slightly outside 0.05 were schooling-cogPGS for fIQ (3%) and WM (5%).

"most of the terms have a substantial part of their distribution overlapping with the ROPE boundaries of 0.02. This means we cannot confirm nor disprove the null hypothesis for a maximal effect size of interest of 0.02 SDs."

Discussion¶

Related Claims¶

• CogPGS and SES each showed large, independent, positive effects on all three intelligence domains — related
• In a European-ancestry subsample, main-effect results stayed generally the same with slightly larger cogPGS and smaller SES effects, and interactions remained non-significant — related
• Relative to chronological age, one year of schooling had its largest ratio on working memory (ratio = 2.2), higher than cIQ (1.1) and fIQ (0.54) — related
• Schooling's effect on general intelligence (g) was 0.136 SD, mirroring the domain findings, with no significant interactions — related
• One year of schooling significantly improves children's crystallized intelligence, fluid intelligence, and working memory (schooling β = 0.13, 0.10, 0.09 respectively) — related
• Within-family cogPGS effect on crystallized intelligence was roughly half the between-family effect, indicating passive genotype-environment correlation — related