Trim-and-fill and funnel graph methods tend to overestimate publication bias.

Evidence

The funnel graph in turn relates to two additional methods of testing for publication bias. The “trim-and-fill” technique detects potentially missing studies by adjusting for funnel plot asymmetry (Duval & Tweedie, 2000a, 2000b; Sutton, Duval, Tweedie, Abrams, & Jones, 2000). The general linear model approach of Vevea and Hedges (1995) focuses on the absence of small studies. It assumes that random effects are distributed normally and that the survival probability of a given study can be described by a stop function around such critical probability points as . and .01. It should be noted, however, that such funnel graph methods tend to overestimate publication bias (Sterne & Egger, 2000). — Thomas F. Pettigrew, Linda R. Tropp, A Meta-Analytic Test of Intergroup…, 48

Categorization

• DEFINE: Defines trim-and-fill and the Vevea-Hedges model, with the caution that funnel methods overestimate bias.
• QUESTION: Worth chasing: whether trim-and-fill’s overestimation applies across fields or specific to this corpus.
• XREF: connects to: Publication bias is unlikely to fully explain the religious priming effect. — Annetta Snell and Miron Zuckerman and Bonnie M. Le, Does Religious Priming Induce Greater…, 88; Industry cherry-picking is continuous with ordinary scientific publication bias — Cailin O’Connor and James Owen Weatherall, The Misinformation Age: How False…, 359; The trim-and-fill adjustment increases the contact-prejudice effect estimate, supporting its robustness. — Thomas F. Pettigrew, Linda R. Tropp, A Meta-Analytic Test of Intergroup…, 107 (+1 more)

Related

• The trim-and-fill adjustment increases the contact-prejudice effect estimate, supporting its robustness. (similarity 0.87)
• Publication bias can be detected through sample-size and effect-size correlations. (similarity 0.84)
• A null correlation between sample size and effect size indicates minimal publication bias. (similarity 0.81)
• Publication bias is unlikely to fully explain the religious priming effect. (similarity 0.80)
• The contact-prejudice effect may be partly an artifact of publication bias. (similarity 0.80)
• Comparing published and unpublished studies tests whether reporting bias inflates effects. (similarity 0.78)
• Meta-analyses overestimate effect sizes because null results go unpublished. (similarity 0.78)