|
Abstract:
|
Statistical learning methods of subgroup identification based on interaction effect are of great interest in the pharmaceutical area. It is desirable to find a subgroup of patients with enhanced treatment effect so that we can efficiently lower the sample size and improve the success rate of drug development projects. Most of the current subgroup identification methods are either optimized for prediction or based on some qualitative criteria. In this presentation, we propose a method called "SQUANT" that integrates the quantitative information into subgroup identification. The new method does not rely on any specific parametric model, and works for continuous, binary and survival response. We will also demonstrate the performance of the proposed method through simulation.
|
