Online Program Home
  My Program

Abstract Details

Activity Number: 403 - Selected Topics on Hypothesis Testing and Statistical Inference
Type: Contributed
Date/Time: Tuesday, August 1, 2017 : 2:00 PM to 3:50 PM
Sponsor: IMS
Abstract #324214
Title: Hypothesis Testing for Simultaneous Variable Clustering and Correlation Network Estimation, with Application to Gene Co-Expression Networks
Author(s): Kevin Lin* and Junwei Lu and Han Liu and Kathryn Roeder
Companies: Carnegie Mellon University, Statistics Department and Princeton University and Princeton University and Carnegie Mellon University
Keywords: correlation graph ; variable clustering ; microarray ; hypothesis test
Abstract:

We develop inferential methods on the difference between correlation networks of two different populations where each node represents a cluster of variables. Previous works have developed the G-latent model to simultaneously cluster variables and estimate the correlation network. This network represents the "supergraph" where edges between nodes represents correlation between two clusters of variables. The motivation of our work is to perform inference between cases' and controls' supergraphs based on microarray expression data. Our work develops three new types of hypothesis tests on these supergraphs which were previously unavailable. First, we can test if an edge exists between two cluster of variables in both the cases' and controls' supergraph. Next, we can test if two variables belong to the same cluster. Lastly, we do a goodness-of-fit test to determine if there is an appropriate number of clusters. We prove the statistical rates and provide a computational method. To demonstrate our methods, we apply them on the CommonMind microarray dataset to infer the differences in the clustered gene co-expression networks between schizophrenic cases and controls.


Authors who are presenting talks have a * after their name.

Back to the full JSM 2017 program

 
 
Copyright © American Statistical Association