Re: st: RE: RE: write all the p values from correlations into a table
Date
Tue, 14 Sep 2010 20:19:28 +0100
Sometimes there are good reasons for collecting multiple P-values, as in
a whole-genome association study (WGAS). However, it is important to
admit that you are collecting multiple P-values, and to use a sensible
multiple-test procedure to define a discovery set of "significant"
P-values which is still convincing, in spite of the huge number of
P-values. (And this discovery set should still be written up for
publication if it is found to be empty, as would have happened with
Tony's example.)
The -qqvalue- and -smileplot- packages, downloadable from SSC, implement
a large number of multiple-test procedures. The issues involved are
discussed at length in Newson et al. (2003), Newson (2003) and Newson
(2010).
I hope this helps.
Best wishes
Roger
References
Newson, R. and the ALSPAC Study Team. 2003. Multiple-test procedures
and smile plots. The Stata Journal 3(2): 109-132. Download from THe
Stata Journal website at
Newson, R. B. Post-parmest peripherals: fvregen, invcise, and qqvalue.
Presented at the 16th UK Stata User Meeting, 9-10 September, 2010.
Download from
I strongly disagree with the philosophy of looking at 600 p-values (and most likely picking off the largest ones). This is poor statistics and poor science. Please read something on multiplicity and why this is bad news.
A recent paper in the Lancet showed that eating cereal in the first trimester increased the probability of a male baby to 0.55. They had done 262 tests on various diets and found the most significant one. A plot of the p-values (done by a critic of the paper) showed a perfect straight line when comparing the p-values with a line of slope 1.
I strongly suggest you abandon this approach.
Tony
Peter A. Lachenbruch
Department of Public Health
Oregon State University
Corvallis, OR 97330
Phone: 541-737-3832
FAX: 541-737-4001
-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Martin Weiss
Sent: Tuesday, September 14, 2010 10:20 AM
To: statalist@hsphsun2.harvard.edu
Subject: st: RE: write all the p values from correlations into a table
<>
***********
capt which estout
if _rc ssc inst estout
sysuse auto, clear
qui ds, has(type numeric)
qui estpost cor `r(varlist)', mat
//show matrix of p-values
mat l e(p)
***********
HTH
Martin
-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu [mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of helen bian
Sent: Dienstag, 14. September 2010 18:17
To: statalist@hsphsun2.harvard.edu
Subject: st: write all the p values from correlations into a table
Dear All,
I want to record the p values only for the correlation between more
than 600 variables.
It is very difficult to record all the p values by the following command
