Table 4.
Model Comparison Among Three Unconditional Piecewise Growth Curves on Cortisol Diurnal Trajectory
| Parameter | Linear spline model | Linear-quadratic spline model | Linear-cubic spline model |
|---|---|---|---|
| Estimate (SE) | Estimate (SE) | Estimate (SE) | |
| Intercept | 9.18 (0.32)*** | 9.18 (0.32)*** | 9.18 (0.32)*** |
| 4-Part piecewise linear model | |||
| CAR slope | 6.32 (0.59)*** | 5.52 (0.49)*** | 6.22 (0.50)*** |
| 1st linear decline slope from 30 min after waking to before lunch | −1.83 (0.08)*** | ||
| 2nd linear decline slope from before lunch to late afternoon | −0.25 (0.04)*** | ||
| 3rd linear decline slope from late afternoon to before bed | 0.07 (0.03)* | ||
| Quadratic and cubic spline model | |||
| Linear decline slope from 30 min after waking throughout the day | −1.68 (0.04)*** | −2.50 (0.08)*** | |
| Quadratic decline slope from 30 min after waking throughout the day | 0.07 (0.00)*** | 0.21 (0.01)*** | |
| Cubic decline slope from 30 min after waking throughout the day | −0.01 (0.00)*** | ||
| REML deviance | 32,959.3 | 33,378.1 | 33,245.7 |
| AIC | 32,993.3 | 33,402.1 | 33,271.7 |
| BIC | 33,045.4 | 33,438.9 | 33,311.5 |
Note: AIC = Akaike information criterion; BIC = Bayesian information criterion; CAR = cortisol awakening response; REML = restricted maximum likelihood. All daily (within-person) covariates were person-mean centered. All between-person covariates were grand-mean centered.
*p ≤ .05, ***p ≤ .001.