Table 4.
Models and model fit statistics
| χ2 | df | CFI | BIC | Model compared to | ΔBIC | |
|---|---|---|---|---|---|---|
| 1. One common factor | 1052 | 77 | 0.71 | 1427.0 | ||
| 2. Three correlated factors (EXT, INT and TP) | 554 | 74 | 0.86 | 969.3 | 1 | 457.7 |
| 3. Four correlated factors (EXT, INT, TP and SUD) | 548 | 71 | 0.86 | 1003.5 | 2 | +34.2 |
| 4. Four factor model (EXT, INT, TP and uncorrelated NDD) | 419 | 71 | 0.90 | 874.5 | 2 | 94.8 |
| 5. Four correlated factors (EXT, INT, TP and NDD; ED on INT) | 385 | 71 | 0.91 | 840.5 | 4 | 34.0 |
| 6. Bifactor model, with four uncorrelated specific factors (ED on INT only) | 400 | 63 | 0.90 | 962.7 | 5 | +122.2 |
| Modified four correlated factors models | ||||||
| 5a. Four correlated factors, BIP on TP and EXT | 384 | 70 | 0.91 | 852.9 | 5 | 12.4 |
| 5b. Four correlated factors, BIP on TP and INT | 383 | 70 | 0.91 | 851.9 | 5 | 11.4 |
| 5c. Four correlated factors, OCD on TP and INT | 385 | 70 | 0.91 | 853.9 | 5 | 13.4 |
| 5d. Four correlated factors, OCD on INT only | 402 | 71 | 0.90 | 857.5 | 5 | 17.0 |
| 5e. Four correlated factors, MDD on INT and TP | 382 | 70 | 0.91 | 850.9 | 5 | 10.4 |
| 5f. Four correlated factors, ED on INT and EXT | model would not run | |||||
| 5g. Four correlated factors, ED on INT and TP | 341 | 70 | 0.92 | 809.9 | 5 | 30.6 |
| 5h. Four correlated factors, ED on TP only | 341 | 71 | 0.92 | 796.5 | 5 | 44.0 |
| 5i. Bifactor model, with four uncorrelated specific factors (ED on TP only) | 366 | 63 | 0.91 | 928.7 | 5h | +132.2 |
CFI – comparative fit index, BIC – Bayesian information criterion, EXT – externalizing factor, INT – internalizing factor, TP – thought problems factor, SUD – substance use disorders factor, NDD – neurodevelopmental disorders factor, BIP – bipolar disorder, OCD – obsessive‐compulsive disorder, MDD – major ‐depressive disorder, ED – eating disorders. The model in bold is the best‐fitting model based on a BIC difference >10. ΔBICs that have plus signs indicate that the more parsimonious models have better fit than the more complex models.