APPENDIX A.
MODEL FIT INDICES FOR AIM 1: HISTORY OF SELF-REPORTED OFFENDING
| Note | |||
|---|---|---|---|
| AIC | BIC | RMSE | |
| Negative Binomiala | 5,131.71 | 5,177.41 | 0.9958 |
| Poissonb | 5,480.38 | 5,521.00 | 1.3647 |
| Hurdle | 5,468.05 | 5,549.29 | 1.4479 |
| Negative Binomial Hurdle | 5,033.19 | 5,119.50 | 1.0434 |
| Poisson Hurdle | 5,468.05 | 5,549.29 | 1.4479 |
| OLS | 5,582.30 | 5,627.62 | 2.7954 |
Model fit indices were averaged across 20 imputed datasets to include missing data. Results indicate that the negative binomial hurdle model had the best model fit.
a
Ordinary count negative binomial model had significantly better fit than the zero-inflated negative binomial model (Vuong test p = .12).
b
Ordinary count poisson model had significantly better fit than the zero-inflated poisson model (Vuong test p = .20).