Group-randomized trials (GRTs) are comparative studies in which investigators randomize identifiable groups to conditions and observe members of those groups to assess the effects of an intervention. They face special problems of design and analysis that must be considered to ensure adequate power and valid results. The most critical problem is that observations taken on members of identifiable groups tend to be positively correlated, thereby violating the independence of errors assumption underlying the common analytic methods. Mixed-model regression methods, generalized estimating equations, and randomization-based methods can provide solutions to this problem if used properly and under the right circumstances. This article provides a brief review of these issues.