Sample size calculations in human electrophysiology (EEG and ERP) studies: A systematic review and recommendations for increased rigor.

Abstract

There is increasing focus across scientific fields on adequate sample sizes to ensure non-biased and reproducible effects. Very few studies, however, report sample size calculations or even the information needed to accurately calculate sample sizes for grants and future research. We systematically reviewed 100 randomly selected clinical human electrophysiology studies from six high impact journals that frequently publish electroencephalography (EEG) and event-related potential (ERP) research to determine the proportion of studies that reported sample size calculations, as well as the proportion of studies reporting the necessary components to complete such calculations. Studies were coded by the two authors blinded to the other's results. Inter-rater reliability was 100% for the sample size calculations and kappa above 0.82 for all other variables. Zero of the 100 studies (0%) reported sample size calculations. 77% utilized repeated-measures designs, yet zero studies (0%) reported the necessary variances and correlations among repeated measures to accurately calculate future sample sizes. Most studies (93%) reported study statistical values (e.g., F or t values). Only 40% reported effect sizes, 56% reported mean values, and 47% reported indices of variance (e.g., standard deviations/standard errors). Absence of such information hinders accurate determination of sample sizes for study design, grant applications, and meta-analyses of research and whether studies were adequately powered to detect effects of interest. Increased focus on sample size calculations, utilization of registered reports, and presenting information detailing sample size calculations and statistics for future researchers are needed and will increase sample size-related scientific rigor in human electrophysiology research.