|
Activity Number:
|
539
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Thursday, August 2, 2007 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Biopharmaceutical Section
|
| Abstract - #310024 |
|
Title:
|
A Mixture Model Approach for Estimating the Proportion of True Null Hypotheses and Adaptive Control of FDR
|
|
Author(s):
|
Ajit Tamhane*+ and Jiaxiao Shi
|
|
Companies:
|
Northwestern University and Northwestern University
|
|
Address:
|
Department of Statistics, Evanston, IL, 60208,
|
|
Keywords:
|
Beta Model ; EM Algorithm ; False Discovery Rate ; Mixture Model ; Normal Model ; Adaptive Control of FDR
|
|
Abstract:
|
We study two mixture distribution models for the distribution of the test statistics with the goal being the estimation of the proportion coming from the null hypothesis and the use of this estimate in adaptively controlling the FDR. The normal model assumes that the test statistics follow a mixture of $N(0,1)$ and $N(\delta,1)$ with $\pi_0$ and $1-\pi_0$ as the mixing proportions. The beta model assumes that the $p$-values follow a mixture of $U[0,1]$ and Beta$(a,b)$. Three methods of estimation are developed for each model. The methods are compared with each other and with Storey's estimator via simulation. Robustness is studied by generating data from other models. The EM algorithm performs best overall when the assumed model holds, but is not very robust to significant model violations. An example is given to illustrate the methods.
|
