regression models for ordinal responses a review of methods

Based on the fit of an unconstrained model model 6

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: umes that the relative risk associated with ‘any laceration’ is equivalent when comparing 4° versus none to 1° – 3° (combined), 3° – 4° (combined) versus none plus 1° – 2° (combined), and so on. In constrast, the CR model assumes that the relative risk associated with ‘any laceration’ is equivalent to 4° versus 3°, 3° – 4° (combined) versus 2°, and so on. The likelihood ratio test of H0: β = 0 is rejected both for the PO and the CR models, implying that midline episiotomy is a strong predictor of lacerations during pregnancy, although both models violated the proportional odds and parallel slopes assumptions (discussed later). The results of fitting a partial proportional odds model to the laceration data are summarized in Table 5a. Based on the fit of an unconstrained model (model 6), the estimated log odds ratio comparing women with ‘any laceration’ to no laceration in relation to midline episiotomy is β (0.7125), whereas the log odds comparing 2° – 4° to none plus 1°, 3° – 4° to none pl...
View Full Document

This document was uploaded on 02/25/2014.

Ask a homework question - tutors are online