Home > Mathematics, Statistics > Vol 11 No 1 June 2023 Abstract 4
Generalized Linear Regression for Ordinal Analysis
Shamita Dutta Gupta

Abstract
When the outcome variable is ordinal, the ordered logit model is a popular choice of the analytical approach. However, often, the data set does not meet the proportional odds assumption. In this research, we propose a generalized linear regression model for the ordinal analysis, which increases the degree of freedom of the model (number of the variable)by a small amount. The proposed model calibrates the “break point” between the ordinal outputs. The proposed model would perform better than the ordered logit model when the proportional odds assumption is not met. It would provide a straightforward approach of linear models with a relatively good fit of calibration data. Another benefit is that the projection with calibrated “breakpoint” is more balanced, that the overrating and underrating count are roughly the same. The projection reflects the original data better at the aggregate level.

Full Text: PDF     DOI: 10.15640/arms.v11n1a4