Table 1.
Fit Indices Comparing Increasing Degrees of Measurement Invariance for the C-SHARP
| Fit index | Configural | Weak | Strong | Strict V1 | Strict V2 |
|---|---|---|---|---|---|
| Chi-square | 4279.1 | 4332.1 | 4451.3 | 4391.7 | 4400.3 |
| df | 2134 | 2180 | 2223 | 2271 | 2286 |
| Chi-square/df | 2.005 | 1.987 | 2.002 | 1.934 | 1.925 |
| p | <0.001 | <0.001 | <0.001 | <0.001 | <0.001 |
| Scaled difference | 47.2 | 167.3 | 54.5 | 19.4 | |
| df | 46 | 43 | 48 | 15 | |
| p | 0.422 | <.001 | 0.240 | 0.196 | |
| CFI | 0.797 | 0.796 | 0.789 | 0.799 | 0.800 |
| ΔCFI | −0.001 | −0.007 | 0.010 | 0.001 | |
| McNCI | 0.189 | 0.188 | 0.177 | 0.192 | 0.193 |
| ΔMcNCI | −0.001 | −0.011 | 0.015 | 0.001 | |
| AIC | 65992 | 65900 | 65814 | 65718 | 65688 |
| RMSEA | 0.056 | 0.055 | 0.056 | 0.054 | 0.054 |
| SRMR | 0.083 | 0.089 | 0.090 | 0.091 | 0.113 |
Models are additive and were tested sequentially, left to right. Constraints for equality across groups placed on each model are as followed: Configural, structure; weak, loadings; strong, intercepts; strict V1, residuals; strict V2, latent variance/covariance. Low values are better for the RMSEA (good: <0.05) and the SRMR (good: <0.08). The AIC is a comparative measure of fit with no intrinsic meaning; lower values indicate better fit. CFI and McNCI are measures of absolute and relative fit; measurement invariance is rejected when ΔCFI <−0.005 or when ΔMcNCI <−0.010. The statistic associated with the preferred model for each index is bolded. Ultimately, the strict V2 model was judged superior and was used as the final model.
C-SHARP, Children's Scale for Hostility and Aggression: Reactive/Proactive; CFI, comparative fit index; McNCI, McDonald's non-centrality index, AIC, Akaike information criterion, RMSEA, root mean square error of approximation, SRMR, standardized root mean residual.