Beyond ANOVA: Comparing Strategies for Analyzing (In)dependent-Sample Means in Small Samples

Read the full article See related articles

Listed in

This article is not in any list yet, why not save it to one of your lists.
Log in to save this article

Abstract

Developmental research often involves comparing cross-sectional or repeatedly measured means in small samples (e.g., N < 100), commonly using ANOVA. Generalized estimating equations (GEE), multilevel modeling (MLM), and structural equation modeling (SEM) are advantageous alternatives, which we didactically explain and demonstrate on N = 42 adolescent boys’ (14–16) alcohol-use data using SPSS and R. To compare their robustness in practice, mixed-design data are simulated under conditions commonly reported in the journal Child Development. Researchers using ANOVA should prefer heteroskedasticity-robust t statistics or CIs to test individual contrasts, whereas MLM and SEM can provide safer omnibus tests, assuming small-sample adjustments are applied to maintain nominal Type I error rates. GEE should be avoided with small samples.

Article activity feed