Estimation and Hypothesis Testing for Four Types of Change Point Regression Models with Non-Thresholded Covariates

Youyi Fong, Fred Hutchinson Cancer Research Institute
Ying Huang, Fred Hutchinson Cancer Center
Peter Gilbert, Fred Hutchinson Cancer Center
Sallie Permar, Duke University Medical Center


Change point models are a diverse set of non-regular models that all depend on change points or thresholds. Many software implementations exist for change point models that are aimed at detecting structural changes in a time series. Motivated by non-time series biometrical applications, the R package \textit{chngpt} provides estimation and hypothesis testing functionalities for four variants of change point regression models that allow covariates not subject to thresholding. We illustrate its usage with a real example from the study of immune responses associated with reduction in mother-to-child-transmission of HIV-1 viruses.