Estimation and Comparative Analysis of Generalized Ordered Logit and Multinomial Logit Models

Abstract

Although both OLOGIT and MLOGIT models are from the logit model family and their mathematical form is similar, these models have many different aspects besides their similarities. In this context, two dependent variables, one nominal and the other ordinal, measuring the same information were placed in the same data set and these models were compared from various perspectives by making these variables dependent variables. These are the significance of the parameters, their suitability for estimation, ease of implementation, and the provision of assumptions. During the implementation, GOLOGIT was put into practice because Ordered Logit did not provide the parallel regression assumption. Although the number of significant parameters is the same in both models, GOLOGIT stands out in terms of providing detailed analysis for each level of each qualitative independent variable and making fewer model estimations than MLOGIT