Under bias-adjusted planning assumptions, approximately 14 independent moderator effect sizes per moderator level are needed for 80% power, and about 60.2% of corpus effect sizes fell below this threshold¶

Claim · All claims
Evidence · 1 study (2 entries) · quant-synthesis r3 · q3
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q3 i? Under the planning scenario (g = 0.38, τ² = 0.24, α = 0.05), achieving 80% power for a two-sided test of the random-effects mean required approximately 14 moderator effect sizes. → Bicakci 2026 q3 i? Approximately 60.2% of the 1,786 moderator effect sizes in the corpus fell below k = 14, indicating limited power is common in STEM meta-analytic research. → Bicakci 2026 (2)

Evidence¶

Bicakci 2026¶

Bicakci, M., Heller, F., Stoeger, H., & Ziegler, A. (2026). From RAMSing to Rigor: Improving Moderator Analysis in STEM Education Meta-Research. Educational Psychology Review. https://doi.org/10.1007/s10648-026-10172-1

q3 · i? · quant-synthesis · r3

Prospective power analysis carried forward from the bias-adjusted RoBMA-PSMA posterior distribution under the stated planning scenario; the article stresses this is a scenario-specific benchmark rather than a fixed rule.

"achieving 80% power for a two-sided test of the random-effects mean was estimated to require approximately 14 moderator effect sizes."

Bicakci 2026 (2)¶

Bicakci, M., Heller, F., Stoeger, H., & Ziegler, A. (2026). From RAMSing to Rigor: Improving Moderator Analysis in STEM Education Meta-Research. Educational Psychology Review. https://doi.org/10.1007/s10648-026-10172-1

q3 · i? · quant-synthesis · r3

Descriptive audit of the same 1,786-effect-size corpus against the k = 14 power benchmark; the article links underpower to unstable estimates and selective reporting patterns.

"approximately 60.2% of the 1,786 moderator effect sizes fell below k = 14, indicating that limited power is common in STEM meta- analytic research."

Discussion¶

Related Claims¶

• Bias-adjusted posterior-predictive benchmarks for STEM moderator effect sizes (0.18, 0.38, 0.64 at the 25th, 50th, 75th percentiles) fall below both Cohen's conventional thresholds and raw empirical quartiles — related
• Guidance, immediate feedback, and system-directed control showed unadjusted moderator signals that did not survive multiplicity adjustment, making them hypothesis-generating only — related