Skip to main content

Table 6 Simulation results in a domain with ICC = 5%, Pearson correlation X1 and random effect 0.0, number of patients 1000, number of centers 50

From: Prediction models for clustered data: comparison of a random intercept and standard regression model

  Apparent performance Test performance
  Standard model Marginal risk calculation Conditional risk calculation Standard model Marginal risk calculation
Harrell’s C-index † 0.79 (0.762; 0.819) 0.79 (0.762; 0.819) 0.82 (0.787; 0.847) 0.78 (0.779; 0.785) 0.78 (0.779; 0.785)
C-index within clusters ‡ 0.80 (0.123) 0.80 (0.123) 0.80 (0.123) 0.79 (0.031) 0.79 (0.031)
Calibration intercept* 0.00 0.03 0.01 -0.01 0.03
Calibration intercept* within clusters ‡ -0.03 (0.455) -0.00 (0.468) 0.01 (0.000) -0.03 (0.492) 0.00 (0.501)
Calibration slope (overall) 1.00 0.96 1.09 0.96 0.92
Calibration slope within clusters ‡ 1.05 (0.095) 1.00 (0.096) 1.10 (0.000) 1.00 (0.027) 0.96 (0.026)
  1. * With calibration slope equal to 1 (i.e. calibration in the large).
  2. † overall performance (2.5 and 97.5 percentiles).
  3. ‡ median of overall performance from 100 simulations (median of within cluster performances).