gologit2

gologit2 is a Stata command for estimating generalized ordered logit models, used when the dependent variable is ordinal. It extends the standard ordered logit by allowing covariate effects to differ across the response thresholds, relaxing the proportional odds assumption that all coefficients are constant across categories.

Estimation is maximum likelihood. The package supports partial proportional odds specifications through an autofit option, which automatically determines which covariates have threshold-specific effects. It also provides likelihood ratio tests for comparing generalized models with the constrained, parallel-lines specification, along with robust standard errors and postestimation tools for predicted probabilities and marginal effects.

Researchers apply gologit2 when the effect of a predictor might change as the outcome moves from one category to the next, such as satisfaction or agreement scales. The command accepts standard Stata data structures, including factor variables and survey weights, and returns coefficients for each threshold where nonparallelism is estimated.

Example usage: gologit2 satisfaction income education, autofit. Options may include testing constraints, customizing reference categories, and selecting output formats for probabilities or margins.

See also: ordered logistic regression; proportional odds model; partial proportional odds model; ologit; ordinal regression.