This preview shows page 1. Sign up to view the full content.
Unformatted text preview: outcomes have received less attention.
However, since all these models are developed under
rather strong assumptions (such as the proportionalodds assumption), departures from these assumptions
may well result in the incorrect model formulation. It is
imperative that the analyst performs (i) goodness-of-fit
tests and an analysis of residuals, and (ii) sensitivity
analysis by fitting and comparing different models.1
Notice that when the number of response levels is
two, all models discussed above reduce to the simple
logistic model for binary responses:
Pr(Y = 1x) = exp (α + x′β)
1 + exp (α + x′β) (13) Our illustration of the models for ordered responses
was based on a single covariate. Our intent here was to 1332 INTERNATIONAL JOURNAL OF EPIDEMIOLOGY FIGURE 1 Relative risks for lacerations by midline episiotomy based on observed data (o), proportional odds (I), constrained ( ) and unconstrained ( ) partial proportional odds models FIGURE 2 Relative risks for lacerations by midline episio...
View Full Document
This document was uploaded on 02/25/2014.
- Spring '11