Table 2. Results from the predictive models tested in this study.
| Models | χ² | df | RMSEA (CI) | CFI | TLI | MDΔχ² | Δdf | ΔRMSEA | ΔCFI | ΔTLI |
|---|---|---|---|---|---|---|---|---|---|---|
| Predictive Models, One Year Lag (Times 1 to 5) | ||||||||||
| Predictive Model (lag 1) | 37110.893* | 24952 | .024 (.024-.025) | .919 | .922 | — | — | — | — | — |
| Invariant Predictive Model (lag 1) | 37647.763* | 25003 | .025 (.024-.025) | .916 | .919 | 461.813* | 51 | +.001 | -.003 | -.003 |
| Predictive Models, Two Year Lag (Times 1–3–5) | ||||||||||
| Predictive Model (lag 2) | 15061.290* | 8774 | .029 (.029-.030) | .938 | .939 | — | — | — | — | — |
| Invariant cross lagged model (lag 2) | 15243.080* | 8791 | .030 (.029-.031) | .937 | .938 | 149.509* | 17 | +.001 | -.001 | -.001 |
| Predictive Models, Four Year Lag (Times 1–5–9) | ||||||||||
| Predictive Model (lag 4) | 14070.069* | 8774 | .027 (.026-.028) | .946 | .947 | — | — | — | — | — |
| Invariant cross lagged model (lag 4) | 13973.726* | 8791 | .027 (.026-.028) | .947 | .948 | 58.165* | 17 | .000 | +.001 | +.001 |
Note. χ² = WLSMV chi square; df = degrees of freedom; RMSEA = Root mean square error of approximation; CI = 90% Confidence Interval for the RMSEA; CFI = Comparative fit index; TLI = Tucker-Lewis index; Δ since previous model; MDΔχ2: chi square difference test based on the Mplus DIFFTEST function for WLSMV estimation. With WLSMV estimation, the χ2 values are not exact, but "estimated" as the closest integer necessary to obtain a correct p-value. This explains why sometimes the χ2 and resulting CFI values can be non-monotonic with model complexity. Given that the MDΔχ2 tends to be oversensitive to sample size and to minor model misspecifications, as the chi-square itself, and to take into account the overall number of MDΔχ2 tests used in this study, the significance level for these tests was set at. 01 [52,53,54].
* p < 0.01.