The results of the PO and ECR models regarding the significant effects of the predictor variables were similar. In both cases, LON, SR, GLSR, and PD were highly significant as predictors of disease severity classes, whereas PC and MAT were not significant (Table 2.1). In both models LON, MAT, and GLSR had negative coefficients, whereas the coefficients for PD, PC, and SR were positive. The ECR model generated slightly smaller standard errors for all of the predictors than the PO model. There was strong evidence of an association with the disease severity for both models (Total P < 0.0001). The value, an index used to measure the
predictive strength of the models, was similar for both models and the likelihood ratio
was higher for the ECR model.