Online Program Home
My Program

Abstract Details

Activity Number: 31 - Categorical Data
Type: Contributed
Date/Time: Sunday, July 29, 2018 : 2:00 PM to 3:50 PM
Sponsor: Biometrics Section
Abstract #328634 Presentation
Title: Optimal Scaling for a Logistic Regression Model with Ordinal Covariates
Author(s): Sanne JW Willems* and Marta Fiocco and Jacqueline J Meulman
Companies: Leiden University and Leiden University and Leiden University & Stanford University
Keywords: Optimal scaling; Nonlinear regression; Logistic regression; Ordinal data; Maximum likelihood

Studies often involve measurement and analysis of categorical data with ordinal category levels. Two methods can be chosen to include ordinal covariates in models. One can use dummy covariates to indicate category memberships. But, since the dummy parameters are estimated independently from each other, the ordering property of the categories may be lost. To keep the ordinal property, one can give integer values to the covariate's categories and include it in the model as a numeric covariate. However, since the data is assumed to be numeric, the property of equal distances between consecutive categories is introduced. The optimal scaling procedure was developed to include ordinal covariates in a proper way. It is applicable to models fitted by least squares estimation. We have developed a method to perform this optimal scaling procedure to models fitted by maximizing a likelihood as well. In a simulation study, the performances of the optimal scaling and the dummy or integer coding were compared for a logistic regression model. Results show that the new method increases the model fit in case of ordinal data, especially in comparison to the integer coding method.

Authors who are presenting talks have a * after their name.

Back to the full JSM 2018 program