Table 2.
Goodness of Fit Criteria
| Step | Purpose | Criteria | |
|---|---|---|---|
| GOF2 | |||
| KMO | 2 | Sampling adequacy | ≥ 0.7 |
| RMSEA | 3 | General fit – model parsimony | ≤ 0.08 [44, 54] |
| SRMR | 3 | Diff. between sample & model | ≤ 0.08 [55] |
| CFI3 | 3 | General fit – Comparative fit | ≥ 0.95 [54, 55] |
| Appropriateness | |||
| Reliability | |||
| Omega | 3 | Internal reliability | ≥ 0.7 [56] |
| CR | 3 | Composite reliability | [57] |
| Validity | |||
| AVE | 3 | Convergent validity | ≥ 0.5 or > CR [58, 59] |
| SC | 3 | Discriminant validity | AVE > SC [57] |
| Corr. with PTSD | 3 | Construct validity4 | p ≤ 0.05 [50, 51, 60] |
1We retained items with loadings between 0.3 and 0.4 when there were strong theoretical reasons for keeping the item [61]
2 Chi-square statistics are not reported as they are not applicable with binary data [44]
3 The Tucker-Lewis index (TLI) is not shown as it is prone to bias and model misspecification with binary data [54]
4 All of the war-related events included, with the exception of inhospitable conditions, meet Criterion A’s definition of traumatic events, demonstrating content validity. Correlations further show that the events we have assumed to be traumatic are, in fact, associated with trauma, demonstrating construct validity