Table 2.
Fit statistics for MMTA models
| Model (Fixed Effects) | AIC | BIC | −2LL | df | LRχ2, dfa |
|---|---|---|---|---|---|
| 1. Intercept, time, MI | 6957.2 | 6965.6 | 6947.2 | 6 | – |
| 2. Added contingency mean effect | 6780.9 | 6791.0 | 6768.9 | 7 | 178.3, 1 |
| 3. Added contingency slope effect | 6722.7 | 6734.5 | 6708.7 | 8 | 60.2, 1 |
| 4. Added interaction between contingency and age | 6717.4 | 6730.9 | 6701.4 | 9 | 7.3, 1 |
| 5. Added mean effect of CS | 6674.5 | 6707.7 | 6674.5 | 10 | 26.9, 1 |
| 6. Added 3-way interaction between contingency, time and age | 6689.9 | 6706.8 | 6669.9 | 11 | 4.6, 1 |
Kenward-Roger correction was used. The only significant random effect was the intercept. Only the factors that contributed statistically significant improvement to model fit are presented. AIC Aikake’s Information Criterion, BIC Bayes Information Criterion, Contingency receiving NS or CS, LL log likelihood, LR likelihood ratio,
afor all LRχ2 tests, the comparison model is the preceding model and p < 0.01