Table 4.
Latent Class Enumeration Across Simulations for the Ordinal Regression Mixture Model.
| Balanced design | Estimated models |
||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| True model |
Selecting two- over one- and three-class |
Selecting three- over two-class |
Percentage of runs not converged |
||||||||||
| N 1 | N 2 | % AIC | % BIC | % aBIC | % BLRT | % AIC | % BIC | % aBIC | % BLRT | One-class | Two-class | Three-class | |
| Balanced (50/50 split) | 3,000 | 3,000 | 76.4 | 5.0 | 50.6 | 73.8 | 22.8 | 0.6 | 0.8 | 23.2 | 0.0 | 1.6 | 71.6 |
| 1,500 | 1,500 | 60.6 | 4.2 | 15.0 | 39.4 | 24.1 | 4.2 | 4.2 | 23.6 | 0.0 | 17.2 | 71.4 | |
| 500 | 500 | 42.0 | 0.4 | 9.2 | 18.6 | 27.4 | 10.8 | 12.6 | 20.6 | 0.0 | 32.2 | 73.0 | |
| 250 | 250 | 34.8 | 0.0 | 15.4 | 17.6 | 27.6 | 14.3 | 18.9 | 19.0 | 0.0 | 36.2 | 69.8 | |
| 100 | 100 | 36.0 | 0.4 | 34.4 | 20.6 | 25.6 | 13.0 | 23.8 | 23.4 | 0.0 | 30.6 | 61.2 | |
| Unbalanced (75/25 split) | 4,500 | 1,500 | 61.8 | 1.0 | 22.4 | 53.8 | 31.1 | 3.7 | 3.7 | 25.0 | 0.0 | 12.2 | 63.0 |
| 2,250 | 750 | 52.2 | 0.0 | 6.8 | 29.4 | 29.3 | 7.3 | 7.6 | 22.0 | 0.0 | 24.0 | 67.4 | |
| 750 | 250 | 40.6 | 0.4 | 6.4 | 18.2 | 24.5 | 10.1 | 10.9 | 18.8 | 0.0 | 32.4 | 73.2 | |
| 375 | 125 | 38.2 | 0.0 | 15.0 | 22.8 | 28.9 | 15.0 | 17.3 | 22.4 | 0.0 | 32.8 | 65.6 | |
| 150 | 50 | 37.0 | 0.4 | 36.0 | 21.4 | 29.4 | 16.4 | 26.1 | 27.2 | 0.0 | 32.6 | 61.0 | |
Note. AIC = Akaike information criterion; aBIC = adjusted bayesian information criterion. % BIC is the percentage of simulations selecting this model using the Bayesian information criteria and BLRT is the bootstrap likelihood ratio test.