Table 2.
Comparison of MLC-D versus MLC-V1 under Simulation Settings II and III. Under either setting, “ML” and “ML-V1” correspond to maximum likelihood estimates under models MLC-D and MLC-V1, respectively. “Bias”, “SD” and “RMSE” represent bias, standard deviation and root mean square error of parameter estimates. Results for 9 selected parameters were displayed, measurement parameters (β11, β12, β13), marginal class prevalences (π1, π2, π3) and ICC parameters (ρ11, ρ22, ρ33). The scale for all numbers is 10−2.
| Method | Quantity | β11 | β12 | β13 | π1 | π2 | π3 | ρ11 | ρ22 | ρ33 |
|---|---|---|---|---|---|---|---|---|---|---|
| Setting II: true model is MLC-D | ||||||||||
| ML | Bias | -0.40 | 0.65 | -0.05 | -0.16 | 0.33 | -0.17 | -0.28 | -0.28 | -0.28 |
| SD | 4.43 | 7.20 | 3.92 | 6.70 | 7.04 | 4.03 | 5.94 | 5.94 | 5.94 | |
| RMSE | 4.45 | 7.23 | 3.92 | 6.70 | 7.05 | 4.03 | 5.94 | 5.94 | 5.94 | |
| ML-V1 | Bias | 0.68 | -0.84 | 0.15 | 0.12 | -0.42 | 0.24 | -1.11 | 5.32 | 0.75 |
| SD | 5.91 | 9.07 | 4.19 | 9.07 | 9.45 | 4.30 | 19.21 | 29.75 | 10.72 | |
| RMSE | 5.95 | 9.11 | 4.19 | 9.07 | 9.46 | 4.31 | 19.24 | 30.22 | 10.74 | |
| Setting III: true model is MLC-V1 | ||||||||||
| ML | Bias | 0.57 | -1.06 | 0.97 | -1.22 | -3.02 | 4.23 | -0.56 | 38.61 | 9.11 |
| SD | 3.77 | 7.53 | 3.52 | 6.51 | 6.35 | 5.00 | 7.48 | 7.48 | 7.48 | |
| RMSE | 3.81 | 7.60 | 3.65 | 6.63 | 7.03 | 6.55 | 7.50 | 39.32 | 11.79 | |
| ML-V1 | bias | -0.70 | 0.94 | 0.01 | -1.44 | -0.83 | 2.27 | -2.42 | 0.70 | -3.63 |
| SD | 4.17 | 7.31 | 4.03 | 5.92 | 5.73 | 6.09 | 11.49 | 12.28 | 10.45 | |
| RMSE | 4.23 | 7.37 | 4.03 | 6.09 | 5.79 | 6.50 | 11.74 | 12.30 | 11.07 | |