Post-Hoc Power Calculations for Null Anxiety/Depression Intervention Trials

What the evidence says

Two meta-analytic sources address whether post-hoc power calculations are used to characterize the reliability of anxiety/depression intervention trials, and they take opposite positions on the practice itself. Zenner (2014) computed a post-hoc power analysis for every included effect size using G*Power, applying an alpha of 0.05 (one-sided) and a target power of 80%. For pre-post effect sizes, the software determined that a sample of n = 41 was needed to detect an effect of d = 0.40; twelve of the included studies met this criterion. For controlled effect sizes, the required sample was n = 78 per group, a bar that only three controlled studies reached. Zenner (2014) also states directly that "many studies are underpowered" and that "there is great heterogeneity," describing the measurement of intervention effects in this literature as "challenging." No further quantification of how many studies fell short, beyond the twelve-of-total and three-of-total figures tied to the specific power thresholds, is given.

Jacob R (2026), by contrast, did not perform a retrospective (post-hoc) power analysis at all, stating that "their usefulness remains controversial" and citing Wang (2010) for that position. This is a methodological decision rather than an empirical finding, so it produces no effect size, direction, or magnitude to report.

Taken together, the evidence on this page establishes that when post-hoc power analysis is applied to intervention trials, a large share of included studies fail to meet the sample-size thresholds needed to reliably detect a moderate effect (d = 0.40) at conventional power. It does not establish how this underpowering relates to the size or direction of measured intervention effects, nor does it isolate anxiety/depression trials specifically from the broader set of studies Zenner (2014) analyzed.

Where the evidence splits

The two sources disagree on whether post-hoc power analysis should be performed at all. Zenner (2014) treats it as a standard quality check and reports specific sample-size thresholds and pass rates. Jacob R (2026) declines to perform the same type of analysis, citing its "controversial" usefulness per Wang (2010). The claims do not state what makes the practice controversial or what design or population difference between the two research contexts drove the different choice — the split is unexplained by the sources on this page beyond the bare citation.

Thresholds and dose-response

Zenner (2014) establishes explicit sample-size thresholds for adequate power at alpha = 0.05 (one-sided) and 80% power, targeting an effect of d = 0.40: n = 41 for pre-post effect sizes, met by twelve studies, and n = 78 per group for controlled effect sizes, met by three studies. No corresponding threshold is given for how power status changes the measured effect size or its direction.

Open empirical questions

• What specifically makes retrospective power analysis "controversial," per the Wang (2010) citation Jacob R (2026) relies on — the claims here don't say.
• What proportion of underpowered studies specifically involve anxiety/depression outcomes, as opposed to the broader mindfulness-intervention set Zenner (2014) analyzed.
• Whether underpowered status is associated with a different measured effect size or direction than adequately powered studies — not addressed in either source.
• Whether the twelve pre-post and three controlled studies that met Zenner's (2014) power thresholds reported different effect sizes than the studies that did not.