Fig. 2From: Risk prediction in multicentre studies when there is confounding by cluster or informative cluster sizeConfounding by Cluster (CBC). Comparison of Basic models (‘Basic’) and models also adjusting for the cluster mean of the exposure (‘Rbar’) fitted by three methods (GLMM, IEE, and GEE with exchangeable correlation.). ICC = 0.2. Performance measures from top left to bottom right: Calibration Intercept, Calibration slope, RPMSE and C-statistic. Conditional (left Panel) and Marginal (right panel) predictions are shown for each measureBack to article page