|
Activity Number:
|
185
- Contributed Poster Presentations: IMS
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Monday, July 29, 2019 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
IMS
|
|
Abstract #302871
|
|
|
Title:
|
Multiple Hypothesis Testing with Discrete Data: Minimally Discrete P-Values
|
|
Author(s):
|
Joshua Habiger*
|
|
Companies:
|
Oklahoma State University
|
|
Keywords:
|
Discrete Data;
False Discovery Rate;
Multiple Testing;
p-value
|
|
Abstract:
|
When testing a null hypothesis with discrete data the natural p-value is conservative, the mid-p-value can be liberal and the randomized p-value is more variable. This work show that this unavoidable ``discreteness effect'' is exacerbated when testing multiple null hypotheses with discrete data. A new ``minimally discrete'' (MD) testing procedure and p-value are developed and shown to dominate traditional approaches for handling the discreteness in that the MD natural p-value is less conservative while the MD randomized p-value is less variable. It is demonstrated that the MD p-values scale up well to high dimensional multiple testing.
|
Authors who are presenting talks have a * after their name.
