Table 3.
Results from the Multivariate Autoregressive Models, Latent Growth Curve Models (LGCM), and Autoregressive Latent Trajectory (ALT) Models.
| Models | χ2 (df) | CM | Δχ2 (df) | CFI | TLI | RMSEA | SRMR | BIC |
|---|---|---|---|---|---|---|---|---|
| 1. Autoregressive model with cross-lagged paths, full model | 76.09 (15)*** | --- | --- | .87 | .61 | .12 | .06 | 5042.91 |
| 2. LGCM, full model a | 45.66 (16)*** | --- | --- | .93 | .78 | .07 | .04 | 7666.11 |
| 3. ALT, full modelb | --- | --- | --- | --- | --- | --- | --- | --- |
| 4. ALT, no slope variancesb | --- | --- | --- | --- | --- | --- | --- | --- |
| 5. ALT, no quadratic variancesb | --- | --- | --- | --- | --- | --- | --- | --- |
| 6. ALT-5 + no slope variancesc | 19.03 (11) | --- | --- | .98 | .91 | .05 | .02 | 7663.52 |
| 7. ALT-6 + no intercept variance for SHSe | 18.66 (15) | 6 | 2.55 (4) | .99 | .97 | .03 | .02 | 7644.72 |
| 8. ALT-6 + no time specific correlations | 34.51 (14)** | 6 | 24.43 (3)*** | .95 | .83 | .07 | .02 | 7655.11 |
| 9. ALT-6 + time specific correlations equalc | 16.41 (13) | 6 | .65 (2) | .99 | .97 | .03 | .02 | 7653.29 |
| 10. ALT-7 + time specific correlations equal | 20.91 (17) | 7 | 2.18 (2) | .99 | .97 | .03 | .02 | 7634.80 |
| 11. ALT-6 + autoregressive paths equal | 16.80 (13) | 6 | 1.09 (2) | .99 | .97 | .03 | .02 | 7654.64 |
| 12. ALT-11 + time specific correlations equal | 19.06 (17) | 11 | 2.46 (4) | 1.00 | .99 | .02 | .03 | 7617.14 |
|
| ||||||||
| ALT-12 model with Predictors | ||||||||
|
| ||||||||
| ALT-12 + Social Support subscales | 25.87 (25) | --- | --- | 1.00 | .99 | .01 | .02 | 7659.68 |
| ALT-12 + Depressed Mood | 19.74 (19) | --- | --- | 1.00 | 1.00 | .01 | .02 | 7626.32 |
| ALT-12 + Perceived Stress | 23.71 (19) | --- | --- | .99 | .97 | .03 | .03 | 7651.74 |
Notes: CM: comparison model in the Δχ2. Δχ2= Satorra-Bentler scaled χ2. Model modifications were made for both SHSe and AFM unless otherwise noted. All models controlled for study group, seasonality, and baseline home smoking ban status. A non-significant χ2 and the following values were indicative of good model fit: > .95 for CFI and TLI, < .05 for RMSEA and SRMR (Geiser, 2013). A BIC change of 10 or more was interpreted as strong evidence for improved model fit (Raftery, 1995). Nested ALT models were compared using the Satorra-Bentler scaled chi-square difference test (Δχ2; Satorra & Bentler, 2001).
Due to error message, residual variance for the observed 12 month SHSe was constrained to 0.
Error message: No convergence.
Error message: PSI not positive definite/negative variance or residual variance.