Masters candidate in Biostatistics, Ziyu Ji, will present:
Plan B Adviser: Lin Zhang
Abstract: In statistical analysis, variable selection is always an important topic that worth to be discussed. Especially, an appropriate variable selection method for group-structured sparse data can play an important role in getting better statistical inference when we are facing high-dimensional genetic problems. Our project is an exploration based on a Bayesian variable selection method, the hierarchical structured variable selection method (HSVS) proposed by Lin Zhang et al. (2014). The objective of this project is to examine the relationship between group size and power/type-I error for both the HSVS and fused HSVS method, and potentially adjust prior in correspondence to group size to achieve best performance in group-level selection. The first part of this study is to test and analyze the effects of group size and other parameter settings on the group selecting results of the HSVS method and fused HSVS method. The second part is providing guidance to correct bias of the methods on simulated datasets by modifying the priors.
Refreshments will be served prior to the presentation.