Table 5.
MLR mean χ2 difference values (Δχ2) and power rates for comparing misspecified complete or partially saturated models with a fixed interaction effect at the within-group level (W0) with the correct models (WIB0, PS_WIBs) without nonlinear effects at the between-group level under conditions of varying numbers of groups (NG) and varying correlation of predictor variables (Φ21).
| Population model: M_WIB0 | ||||||
|---|---|---|---|---|---|---|
| Misspecified Analysis Models at the Within-Group Level | ||||||
| W0B0 vs. WIB0 | PS_W0Bs vs. PS_WIBs | |||||
| NG | Δχ2 | Δdf | Power (%) | Δχ2 | Δdf | Power (%) |
| Φ21 = 0 | ||||||
| 200 | 180.22 | 1 | 100 | 183.61 | 1 | 100 |
| 500 | 439.28 | 1 | 100 | 455.09 | 1 | 100 |
| 1000 | 872.00 | 1 | 100 | 909.39 | 1 | 100 |
| Φ21 = 0.30 | ||||||
| 200 | 212.27 | 1 | 100 | 215.98 | 1 | 100 |
| 500 | 518.49 | 1 | 100 | 536.94 | 1 | 100 |
| 1000 | 1026.03 | 1 | 100 | 1070.55 | 1 | 100 |
Δ χ2 is the mean of the Monte Carlo χ2 difference values. Misspecified models with either an interaction effect added to a linear model or an existing interaction effect fixed to zero are in italics.