A0406
Title: A global kernel estimator for partially linear varying coefficient additive hazards models
Authors: Hoi Min Ng - The Hong Kong Polytechnic University (Hong Kong) [presenting]
Kin Yau Wong - Hong Kong Polytechnic University (Hong Kong)
Abstract: Recent technological advances have made it possible to measure different types of omics features on a large number of subjects. In the studies of chronic diseases such as cancer, it is of great interest to integrate different types of omics features to build a comprehensive understanding of the disease mechanisms. Despite extensive studies on integrative analysis, it remains an ongoing challenge to model the interaction effects among types of omics features due to heterogeneity across data types. A flexible semiparametric varying coefficient additive hazards model that allows for such interaction effects is developed. The additive hazards model is considered due to its simple interpretation and availability of closed-form estimators. It is noted that most existing kernel-smoothing methods for semiparametric varying-coefficient models are inefficient, as they effectively treat the baseline hazard function as changing with the varying-coefficient variable in the estimation, whereas in the model formulation, the baseline hazard function does not vary. A novel kernel-smoothing method is proposed for estimation which makes use of the fact that the baseline hazard function is shared. The theoretical properties of the proposed estimators are established and their finite-sample performance is investigated by large-scale simulation studies. Applications are provided for a motivating cancer genomic study.
