Generalized Ordered Logit/Partial Proportional Odds Models for Ordinal Dependent Variables

Abstract

This article describes the gologit2 program for generalized ordered logit models. gologit2 is inspired by Vincent Fu's gologit routine (Stata Technical Bulletin Reprints 8: 160-164) and is backward compatible with it but offers several additional powerful options. A major strength of gologit2 is that it can fit three special cases of the generalized model: the proportional odds/parallel-lines model, the partial proportional odds model, and the logistic regression model. Hence, gologit2 can fit models that are less restrictive than the parallel-lines models fitted by ologit (whose assumptions are often violated) but more parsimonious and interpretable than those fitted by a nonordinal method, such as multinomial logistic regression (i.e., mlogit). Other key advantages of gologit2 include support for linear constraints, survey data estimation, and the computation of estimated probabilities via the predict command.

Baldness and the androgen receptor: the AR polyglycine repeat polymorphism does not confer susceptibility to androgenetic alopecia

“…All models were fitted using the statistical package Stata v9 (StataCorp 2005, http://www.stata.com), and the Stata add-on gologit2 (Williams 2006) was used extensively.…”

The impact of smoking and polymorphic enzymes of xenobiotic metabolism on the stage of bladder tumors: a generalized ordered logistic regression analysis

“…In this section, our objective is to use the generalized ordered logistic model to study the impact of smoking, age, and NAT2 genotypes on the progression of bladder tumors for the same clinical data of Tunisian patients. The motivation for using the generalized ordered logistic procedure which was introduced by Williams (Williams 2006) is that it can fit three special cases of generalized models: the proportional odds/parallel-lines model, the partial proportional odds model, and the logistic regression model.…”

“…In our Stata program, the gologit procedure goes through an iterative process by fitting an unconstrained model and then we obtain a series of Wald tests on each variable to see whether the coefficients differ across equations as explained by Williams (2006). If the Wald test is statistically insignificant for one or more variables, the variable with the least significant value on the Wald test is constrained to have equal effects across equations.…”

Workplace smoking restrictions: smoking behavior and the intention to change among continuing smokers

“…To verify this, the Brant test (Brant 1990) was used. In case of violation of the parallel regression assumption, generalized, ordered logistic regression analysis, as implemented in the STATA procedure ''gologit2'', was used to fit partial proportional odds models (Williams 2006). Thus, for covariates that did not meet the proportional odds assumption, two different regression coefficients were estimated.…”