This preview shows page 1. Sign up to view the full content.
Unformatted text preview: dds
model performs better than the constrained model.
Similarly, the fit of a continuation-ratio model was
compared to the observed data (Figure 2). Once again,
the continuation ratio model does not adequately fit
the data for the first continuation-ratio (relative
risk = 2.7), whereas the model fits well for the other
three continuation-ratios. On the other hand, when the
ordering of the response variable is subjective, then a
polytomous logistic model is more preferable than
fitting (k – 1) simple logistic models. If the ordering is
valid, then an appropriate ordinal model must be chosen
from the class of ordinal models, the choice made based
on the goals of statistical analyses.
Finally, Koch et al.20 developed a two-stage procedure called as Functional Assymptotic Regression
Methodology (FARM) for fitting a partial proportional
odds model based on the weighted least squares estimation procedure. Although we did not consider fitting this model, the interested reader is referred to the
papers by Koch et al.,20 and Peterson and Harrell6 for
thorough review and discussion. CONCLUSIONS
This paper presents a synthesized review of generalized
linear regression models for analysing ordered responses. The cumulative logit and the continuationratio models for ordinal responses have been the
primary focus in epidemiological and biomedical applications,1–10,15,19–23 while other models for the analysis of ordinal...
View Full Document
This document was uploaded on 02/25/2014.
- Spring '11