Skip to main content

Advertisement

Table 2 Results from the prediction models for data simulated with prevalence = 50%, ICC = .20, n = 5000 (J = S = 50, nj = nc=100), β1 = 1.5 (the default setting)

From: Prediction models for clustered data with informative priors for the random effects: a simulation study

  Optimal score FREQ BAYES.WI BAYES.LI BAYES.MI BAYES.HI FREQ.2 FREQ.3 FREQ.5
Overall Brier score 0 .191 .192 .179 .174 .170 .173 .170 .167
Overall C-index/AUC 1 .782 .781 .808 .818 .826 .822 .827 .833
Overall calibration slope 1 .911 .907 .957 .982 .989 .965 .972 .994
Within cluster C-index/AUCa 1 .805 [.037] .805 [.037] .805 [.037] .805 [.037] .805 [.037] .805 [.037] .805 [.037] .805 [.037]
Within cluster calibration slopea 1 .914 [.102] .914 [.102] .947 [.091] .956 [.078] .963 [.058] .954 [.092] .973 [.080] .977 [.062]
  1. amean[sd]