UCLA

Thresholding variable plays a crucial role in subgroup identification for personalizedmedicine. Most existing partitioning methods split the sample basedon one predictor variable. In this paper, we consider setting the splitting rulefrom a combination of multivariate predictors, such as the latent factors, principlecomponents, and weighted sum of predictors. Such a subgrouping methodmay lead to more meaningful partitioning of the population than using a singlevariable. In addition, our method is based on a change point regression modeland thus yields straight forward model-based prediction results. After choosinga particular thresholding variable form, we apply a two-stage multiple changepoint detection method to determine the subgroups and estimate the regressionparameters. We show that our approach can produce two or more subgroupsfrom the multiple change points and identify the true grouping with high probability.In addition, our estimation results enjoy oracle properties. We design asimulation study to compare performances of our proposed and existing methodsand apply them to analyze data sets from a Scleroderma trial and a breastcancer study.