From: Does group-based trajectory modeling estimate spurious trajectories?
Criteriaa | Scenario 1: Three distinct trajectory subgroups | Validity of classification | ||||||
1 | 2 | 3 | ||||||
Average posterior probability | 1.00 | 0.99 | 0.98 | All criteria suggest good classification | ||||
Mismatch | 0.03 | 0.23 | -0.26 | |||||
Relative entropy | 0.98 | |||||||
Scenario 2: Different range of values of Y values across subgroups | ||||||||
Average posterior probability | 1.00 | 1.00 | All criteria suggest good classification | |||||
Mismatch | 0.00004 | -0.00004 | ||||||
Relative entropy | 1.00 | |||||||
Scenario 3: Time point-specific overlap in the distribution of Y | ||||||||
Average posterior probability | 0.93 | 0.56 | 0.92 | All criteria suggest that the classification is not optimal | ||||
Mismatch | 26.44 | -35.27 | 8.82 | |||||
Relative entropy | 0.34 | |||||||
Scenario 4: Increasing subgroup with variance | ||||||||
1 | 2 | 3 | 4 | 5 | 6 | |||
Average posterior probability | 0.91 | 0.89 | 0.90 | 0.86 | 0.91 | 0.88 | All criteria suggest good classification | |
Mismatch | 0.19 | 0.05 | 0.23 | -0.89 | 0.49 | -0.07 | ||
Relative entropy | 0.87 | |||||||
Scenario 5: Rainbow effect | ||||||||
1 | 2 | 3 | ||||||
Average posterior probability | 0.84 | 0.83 | 0.84 | Mismatch and entropy suggest that the classification is not optimal | ||||
Mismatch | 1.30 | -2.89 | 1.59 | |||||
Relative entropy | 0.65 | |||||||
Scenario 6: No temporal patterns | ||||||||
1 | 2 | 3 | 4 | |||||
Average posterior probability | 0.89 | 0.89 | 0.84 | 0.90 | Mismatch and entropy suggest that the classification is not optimal | |||
Mismatch | 0.52 | -1.97 | 0.45 | 1.00 | ||||
Relative entropy | 0.66 |