Impact of Small Group Size on Neighborhood Influences in Multilevel Models

Objective: Although there is a growing body of literature on sample size in multilevel or hierarchical modeling, few studies have examined the impact of group size < 5. Design: We examined the impact of a group size less than five on both a continuous and dichotomous outcome in a simple two-level multilevel model utilizing data from two studies. Setting: Models with balanced and unbalanced data of group sizes 2 to 5 were compared to models with complete data. Impact on both fixed and random components were examined. Results: Random components, particularly group-level variance estimates, were more affected by small group size than were fixed components. Both fixed and random standard error estimates were inflated with small group size. Datasets where there are a large number of groups yet all the groups are of very small size may fail to find or even consider a group-level effect when one may exist and also may be under-powered to detect fixed effects. Conclusions: Researchers working with multilevel study designs should be aware of the potential impact of small group size when a large proportion of groups has very small (< 5) sample sizes.