MSCOLOGIT: Stata module to estimate multiscale ordered logit model

mscologit fits the multiscale ordered logit model that expresses the association between covariates indepvars and an ordinal outcome variable. mscologit is a generalization of the standard proportional odds ordered logit model, but unlike the standard model, mscologit accommodates the presence of multiple response scales to measure the same rating dimension in the data, as when response formats for the same rating question differ across surveys or survey waves. Alternative outcome variables are listed as depvars, but typically there is but one valid observation in depvars for each row of data (e.g., because each survey respondent had been presented with one particular response format in a particular survey, country, survey wave, or sample split). mscologit is a derivative of the proportional odds model, and is sharing all features of this ordered logit model. mscologit permits to relax the parallel lines assumption of the proportional odds model through its lo and up options, whereby a multiscale variant of the generalized ordered logit model (Fu 1998, Williams 2006, 2016, Fullerton 2009, Agresti 2010, Fullerton and Xu 2016) may be estimated. As a wrapper for meglm, gllamm, and runmlwin, mscologit also permits to implement multilevel specifications of the multiscale ordered logit model; in the current implementation, up to 5 hierarchical levels may be specified in mscologit.