A Simulation Study of Properties of a Regressive Logistic Model

Document Type


Publication Date



A regressive logistic model for the analysis of data with dependent binary observations is constructed by successively conditioning on preceding observations. The properties of this model are investigated and compared to those of the ordinary logistic regression model in which the dependence is not considered, using computer simulation. Comparison criteria include the magnitude of the bias and the total mean square error (MSE) of the regression coefficient (β) and the significance level. The results suggest the regressive model significantly improves the estimation of the regression coefficient. © 1989, Taylor & Francis Group, LLC. All rights reserved.

This document is currently not available here.