Sample size critically shapes the reliability of EEG case-control findings in psychiatry

Electroencephalography (EEG) studies in psychiatry have produced highly inconsistent findings, particularly in case-control designs. Small sample sizes are widely assumed to contribute to this variability, yet their impact on the reliability of EEG group comparisons has not been systematically quantified at scale. Here, we address this key gap using a large multisite resting-state EEG dataset comprising 2,874 participants aged 5-18 years, including individuals with attention-deficit/hyperactivity disorder, autism spectrum disorder, anxiety, learning disorders, and healthy controls. We extracted a comprehensive set of spectral, temporal, and complexity EEG features and performed repeated case-control comparisons across a wide range of sample sizes using extensive random subsampling. Results from small samples were unstable, with inflated and highly variable effect sizes across iterations. By contrast, larger samples produced consistent and reproducible findings, converging on uniformly small but robust effects. Statistical power rose steeply with sample size, whereas false positive rates remained relatively stable. These results demonstrate the central role of sample size in shaping EEG study outcomes and challenge the utility of conventional case-control approaches for biomarker discovery in psychiatry.