Table 3

Model adequacy assessments of latent trajectory class models based on different assumptions for K=5 classes, by gender in the NIH-AARP cohort

ModelDescriptionBICProportion per class %Average posterior probability assignmentRelative
entropy (EK)
Degree of separation (DoSK)
Men
 AHomoscedastic347632251: 22: 21: 6: 183: 84: 85: 91: 950.780.10
 BHeteroscedastic351164641: 26: 20: 7: 685: 82: 84: 89: 86-0.15
 CRandom intercept336485668: 24: 4: 3: 290: 82: 89: 85: 820.810.34
 DRandom slope332546363: 19: 13: 3: 472: 70: 74: 83: 830.590.26
 ERandom quadratic, Equal *
 FRandom quadratic, Proportional 3301301  68: 25: 4: 3: 0.481: 74: 87: 83: 740.680.36
 GRandom quadratic, Unrestricted332000556: 30: 8: 5: 0.474: 70: 80: 79: 880.630.33
Women
 AHomoscedastic228950945: 36: 13: 4: 190: 84: 88: 93: 960.830.14
 BHeteroscedastic223844432: 30: 15: 15: 887: 83: 85: 89: 91-0.05
 CRandom intercept224068152: 35: 9: 2: 289: 82: 86: 89: 910.790.21
 DRandom slope219315579: 12: 5: 2: 291: 80: 79: 86: 910.820.34
 ERandom quadratic, Equal218822495: 5: 0: 0: 049: 88: 0: 0: 00.260.05
 FRandom quadratic, Proportional 2169793 41: 33: 21: 3: 274: 79: 80: 83:840.660.09
 GRandom quadratic, Unrestricted218770767: 23: 6: 3: 184: 77: 84:82: 870.730.34
  • *Failed to converge.

  • BIC, Bayesian information criteria; NIH, National Institutes of Health.

  • Bold values are the chosen according to the lowest value.