How many clusters do I need to fit a multilevel model?

In this edition of CBA Office Hours, Dan discusses a question that frequently comes up in our multilevel modeling workshop, namely, “How many clusters do I need to be able to fit a multilevel model?”  Here, clusters refers to upper-level units, so in the case of individuals nested within groups, the groups, and in the case of repeated measures, the individuals studied over time.  The discussion centers on how to appropriately estimate and obtain inferences from a multilevel model with as few as 10 clusters.

For additional information on this issue, see the excellent review paper recently published by Dan McNeish and Laura Stapleton in Educational Psychology Review.

Note: This video is an expanded remake of an earlier version on the same topic.

Related Articles