|
Activity Number:
|
92
- Section on Risk Analysis P.M. Roundtable Discussion
|
|
Type:
|
Roundtables
|
|
Date/Time:
|
Monday, August 3, 2020 : 12:00 PM to 1:00 PM
|
|
Sponsor:
|
Section on Risk Analysis
|
|
Abstract #309690
|
|
|
Title:
|
Is a Risk Estimate from a Meta-Analysis Reliable?
|
|
Author(s):
|
Stanley Young*
|
|
Companies:
|
CGStat
|
|
Keywords:
|
observational study;
meta-analysis;
p-value;
reliability
|
|
Abstract:
|
In risk analysis where a research question/claim is at issue, it is common to gather scientific papers that address the question/claim and put results from these base papers – usually risk ratios or odds ratios and their confidence limits – into a meta-analysis. Where the base papers are observational in nature, the reliability of the papers needs to be examined – which can be done via simple counting to estimate the analysis search space. It is very often the case that there is large heterogeneity across the base papers. The usual meta-analysis computes a weighted average across the base papers. We observe that base papers tend to have large analysis search spaces and results drawn from them often form a two-component mixture; some papers have small p-values while others appear random. Averaging over a mixture does not make sense. This roundtable is devoted to the question of the reliability of meta-analysis studies where the base papers are observational. Examples are given and open questions are posed.
|
Authors who are presenting talks have a * after their name.
Back to the full JSM 2020 program
|
