Exploratory CART for Semi-Markov Models,
BROWN UNIV PROVIDENCE RI
Pagination or Media Count:
INITIALLY This paper describes a nonparametric application of CART Breiman et al., 1984 to semi-Markov models, to provide a nonparametric regression analysis of transition data. Modeling data without any assumptions about the nature of the underlying distributions is needed for investigating predictor effects in an exploratory analysis. The semi-Markov assumption specifies a structure for the transition process, which is characterized by the one-step transition distributions. The nonparametric regression is done on these distributions. For each one-step transition distribution, the recursive partitioning of the variable space allows greater interpretability of the data by splitting the data into homogeneous subpopulations, and by providing insight into the relative importance of the different predictors, and the way in which they interact. This method is then applied to modeling payment source changes of nursing home residents.
- Statistics and Probability