Defining and estimating effects in cluster randomized trials: A methods comparison