Unit of Analysis Error

Unit of Analysis Error

 

When investigators use any sort of cluster randomization (randomize by physician instead of patient, practice instead of physician or patient, or village instead of participant) and analyze as if they have randomized according to patient or participant, they have made a unit of analysis error. The appropriate analysis acknowledges the cluster randomization, and takes into account the extent to which outcomes differ between clusters independent of treatment effect.

« Back to Glossary Index