In my experience, the need for extra decimal places in P-values is
usually motivated by multiple comparisons. For instance, if you are
doing a genome scan of 10,000 polymorphism associations, then, on
average, one of these polymorphism associations will have a sample
P-value of P<=0.0001, even if all corresponding population associations
are zero.
In these cases, there is not yet a consensus, even amongst
statisticians, about what constitutes a "conventional significance
test". Most of the seminal references on practical multiple-test
procedures for large sets of P-values are dated 2001 or later. Also, it
would be very tedious to look at the returned results one by one,
especially as, with e-class commands, you must either do some extra
work, or get -estout-, -outreg- or -parmest- to do it for you.
Best wishes
Roger
Roger Newson
Lecturer in Medical Statistics
Respiratory Epidemiology and Public Health Group
National Heart and Lung Institute
Imperial College London
Royal Brompton campus
Room 33, Emmanuel Kaye Building
1B Manresa Road
London SW3 6LR
UNITED KINGDOM
Tel: +44 (0)20 7352 8121 ext 3381
Fax: +44 (0)20 7351 8322
Email: [email protected]
Web page: www.imperial.ac.uk/nhli/r.newson/
Departmental Web page:
http://www1.imperial.ac.uk/medicine/about/divisions/nhli/respiration/pop
genetics/reph/
Opinions expressed are those of the author, not of the institution.
-----Original Message-----
From: [email protected]
[mailto:[email protected]] On Behalf Of n j cox
Sent: 21 September 2007 19:31
To: [email protected]
Subject: Re: st: RE: exact (more decimals) p values ?
There is a more general point which might get obscured
by references to -parmest-, -estout- and so forth.
It is that you can see more decimals by looking at
the saved results, using -return list- or -ereturn list-
as appropriate. In some cases the P-value is one
such saved result; in others you have do more work
to get at a P-value.
Programs such as -parmest- and -estout- do their magic
by seeking out such returned results.
A different kind of point is that once a P-value shows
that you are far out in the tail of a sampling distribution,
the accuracy (let alone exactness) of that P-value is
especially sensitive to whether the underlying
assumptions are indeed satisfied. Scientifically
it makes no difference to me, and no doubt at least
some others, what follows P = 0.0000,
as the only issues are then _interpreting_ a model
fit that passes every conventional significance
test.
Newson, Roger B
I personally aencounter this problem all the time. My usual solution is
to use the -parmest- package, downloadable from SSC, which produces an
output dataset (or resultsset) with one observation per parameter and
data on estimates, confidence intervals, P-values and other parameter
attributes. This dataset can be listed to the log and/or saved to disk
and/or written to memory, overwriting the existing dataset.
For instance, you might type
regress mpg weight foreign
parmest, list(parm estimate stderr t min96 max95 p) format(p %-8.2g)
and get the results listed with the P-value in a left-justified format,
usually with 2 significant figures. The on-line help for -parmest-
contains many more references to other tricks that can be done with
-parmest-, together with other packages on SSC. So does my website (see
my signature below).
Alternatively, there is probably a solution with either -estout- or
-outreg-, which you can also download from SSC.
Marcus Fischer
is there a way to report p values with more than 4 decimals ?
(in case of low p values STATA reports 0.0000). I would like to know the
exact p value.
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
