regression models for ordinal responses a review of methods

In reality dichotomization can be somewhat arbitrary

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: seline category (i.e. women with no laceration). DISCUSSION Choice of an Ordinal Model The choice between the cumulative logit and continuation-ratio models merit further discussion. Armstrong and Sloan15 argue that when the cumulative logit model is valid implying that the cumulative logodds ratios are a constant, say β*, then the continuationratio model will begin at βj = β*, but will approach 0 as j increases. This argument has led to the cumulative logit model being proposed as an adjunct to the Cox’s proportional hazards model13 i.e. continuation-ratio model for survival data when hazard rates of groups are thought likely to converge with time (see McCullagh4 and McCullagh and Nelder14 for a more thorough discussion). The choice of a simple logistic model to an ordered response has its own disadvantages. Based on extensive simulations, Armstrong and Sloan15 conclude that the logistic model attains only between 50 and 75% efficiency, relative to the cumulative logit model for a fivelevel ordered response. Moreover, efficiency...
View Full Document

Ask a homework question - tutors are online