Mixed-effects meta-analytic models trade precision for fewer false positives

Evidence

A mixed-effects model was used in the current analyses, where it is assumed that there are systematic differences in the effect sizes that can be accounted for by moderating variables, but allows for the possibility that other, unmeasured factors may be influencing the variance of the effect sizes (Raudenbush, 1994). A mixed-model approach was a more conservative choice and less prone to Type I errors compared to the fixed-effects approach (Hedges & Vevea, 1998). Each correlation coefficient was transformed using Fisher’s Zr transformation (Hedges & 01kin, 1985) and was then weighted for differences in sample sizes. — Blake M. Riek, Eric W. Mania, and Samuel L. Gaertner, Intergroup Threat and Outgroup Attitudes, 79

Categorization

• XREF: Connects to the reader’s empirical backbone work — the choice of mixed- vs. fixed-effects models is a key methodological decision for citing effect sizes from this review. — connects to: Group esteem threat effects may reflect publication bias — Blake M. Riek, Eric W. Mania, and Samuel L. Gaertner, Intergroup Threat and Outgroup…, 94; Experimentally manipulated realistic threat shows far weaker attitude effects than measured threat — Blake M. Riek, Eric W. Mania, and Samuel L. Gaertner, Intergroup Threat and Outgroup…, 119; Some false positives during threshold testing indicate healthy response bias — Jennifer Lentz, Psychoacoustics- Perception of Normal…, 347 (+1 more)

Related

• Group esteem threat effects may reflect publication bias (similarity 0.85)
• Experimentally manipulated realistic threat shows far weaker attitude effects than measured threat (similarity 0.82)
• Meta-analytic reviews must exclude regression-only studies because effect sizes require zero-order correlations (similarity 0.81)
• Broad multi-facet threat scales capture stronger threat-attitude correlations than single-facet ones (similarity 0.80)
• Group threats act through intergroup emotions to reach outgroup attitudes. (similarity 0.79)

Broader Connections

• Some false positives during threshold testing indicate healthy response bias #corroborates (0.75) — Jennifer Lentz, Psychoacoustics- Perception of Normal…, 347