Abstract
The prognostic factors affecting survival time are important components of survival analysis. Covariates may be evaluated by the tree-structured method, but this method has high-dimensional interaction. Tibshirani and LeBranc (1993) proposed an automatic binary logistic Estimator (ABLE) method for exploring such indices in binary outcomes. ABLE is constructed from linear combinations of index functions, and the regression parameters are estimated from maximum likelihood in a logistic regression model framework, namely, the automatic proportional hazard Estimator (APHE) model. In this paper, ABLE is extended to the right-censored survival response. The model incorporates the survival multivariate adaptive regression spline algorithm (LeBranc and Crowley, 1999). The effectiveness of the APHE method was validated on real data from phase III clinical trials for infectious diseases and via small-scale simulations.
