We know (from Agresti and many others) that some exact tests (e.g.,
Fisher's test and exact logistic regression) are not very accurate, so
I hesitate to recommend this approach in general.
Frank
Frank E Harrell Jr Professor and Chairman School of Medicine
Department of Biostatistics Vanderbilt University
On Mon, 16 Aug 2010, David Judkins wrote:
I have never heard of anyone working on this issue, but it sounds very
interesting. Paul Flyer has been very active in urging the use of
exact rerandomization tests in the analysis of clinical trials. Others in the
area include Gary Koch and Lisa LaVange. However, I do
not think any of them have worked with multiply imputed data. The only
plausible person I can think of to have worked on this issue is
Paul Rosenbaum. He had an article in Statistical Science a few years back on
how to exploit covariate information within a
randomization-based inferential procedure.
David Judkins
Senior Scientist
Westat
1650 Research Boulevard
Rockville, MD 20850
(301) 315-5970
[email protected]
From: Impute -- Imputations in Data Analysis
[mailto:[email protected]] On Behalf Of Juned Siddique
Sent: Wednesday, August 11, 2010 4:24 PM
To: [email protected]
Subject: Combining Exact Tests Across Multiply Imputed Datasets
Hi.
I am looking for guidance on combining the results of exacts tests on multiply
imputed datasets. Any help would be appreciated. Thank
you.
Sincerely,
Juned Siddique
