Change-Point Models in Logistic Regression

In this paper we consider a change-point logistic regression model on hazard rates. We estimate both the logistic regression parameters and the change-point using the maximum loglikelihood function. The maximum likelihood estimation of logistic regression parameters for any fixed change-point are obtained by the Newton-Raphson iterative algorithm. As a byproduct of this algorithm we can also obtain the covariance matrix of estimated parameters. We discuss a model selection procedure based on the Schwarz information criterion. We also checked model goodness-of-fit in the respect of signed deviance residuals. The suggested change-point logistic model was explained through a real data set of head-and-neck cancer survival times. The proposed change-point logistic regression model performed quite well compared to a single logistic model with no change-point.