regression models for ordinal responses a review of methods

Tests for model assumptions could also be viewed as

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: is based on the assumption that the dichotomization for the logistic model is made close to the optimal point. In reality, dichotomization can be somewhat arbitrary 1331 REGRESSION MODELS FOR ORDINAL RESPONSES which may violate the underlying assumptions of the model. The arbitrariness is worsened in situations when a logistic model is fit to a response that has many ordered categories. A more intuitive choice between the proportional odds and continuation-ratio models can be based on the goals of the statistical analysis. Assuming that both models are valid, if an a priori interest is to estimate the risk of 4° laceration relative to other groups (none and 1° – 3° combined), then the cumulative logit model is the obvious model choice. On the contrary, if the analyst is interested in estimating the risk by comparing 4° laceration to 3°, then the continuation-ratio model is the preferred model. In general, the choice of a model depends on how the logits are formulated, a priori. The analyst should, however, be wary of departures from the underlying model assumptions (proportional odds and parallel slopes assu...
View Full Document

This document was uploaded on 02/25/2014.

Ask a homework question - tutors are online