Skip to main content

Table 5 Results of the association parameter α estimated from the TVCM and joint model fitted to data generated considering slight and gross misspecifications of the longitudinal trajectories (scenarios 7 and 8 of Table 1), σε(0.1,0.3,0.5) with CV (3.1%,9.4%,15.6%) and the true α=0.3

From: Joint model robustness compared with the time-varying covariate Cox model to evaluate the association between a longitudinal marker and a time-to-event endpoint

  Slight misspecificationGross misspecification
σεModelEstESEASE%BiasCPEstESEASE%BiasCP
0.1TVCM(1x/week)0.2950.0560.056-1.7950.2970.0550.056-1.095
 TVCM(4x/week)0.3090.0540.0553.0960.3100.0530.0543.396
 joint model0.2970.2090.060-1.0920.2800.2220.059-6.791
0.3TVCM(1x/week)0.2770.0540.054-7.7930.2800.0530.054-6.793
 TVCM(4x/week)0.2920.0510.053-2.7950.2940.0510.052-2.095
 joint model0.2840.2910.067-5.3910.273*0.2900.059-9.091
0.5TVCM(1x/week)0.2490.0500.051-17830.2520.0500.051-1684
 TVCM(4x/week)0.2640.0470.050-12890.2660.0470.049-1190
 joint model0.2650.1920.071-12910.265*0.1670.065-1291
  1. Mean of the maximum likelihood estimates (Est), empirical Monte Carlo standard error (ESE), asymptotic standard error (ASE), percentage bias (%Bias) and 95% coverage probabilities (CP) are shown. The star (*) indicates that one extreme outlier estimate was removed