EcoSta 2019: Start Registration
View Submission - EcoSta2019
Title: Regularized variable selection for high dimensional survival data with unknown link function Authors:  Haitao Zheng - Southwest Jiaotong University (China) [presenting]
Abstract: Aft model is one of the most commonly used models to handle survival data. However, the model assumption may not be correct in the real. We do not make any assumption on link function of the model. We use kernel estimation method to estimate the unknown link function. Then, we use a weighted least squares method with censoring constraints and sparse penalization to select high-dimensional covariates. In simulation studies, we compare the tpr, tnr, tdr and tndr of the proposed model with aft model and the neural network model from R package vsurf to illustrate the performance of each method.