Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: st: RE: Can I have different results from Stata13 depending on the OS?
From
Daniel Feenberg <[email protected]>
To
"[email protected]" <[email protected]>
Subject
Re: st: RE: Can I have different results from Stata13 depending on the OS?
Date
Tue, 21 Jan 2014 16:41:37 -0500 (EST)
On Tue, 21 Jan 2014, Stas Kolenikov wrote:
In my experience, there were tiny differences in -summarize- results
between some of Stata versions obtained by experimenting with -set
processors #- on the order of about 100*c(epsdouble) (i.e., way more
than c(epsdouble), and enough to throw -assert r(mean) == #- off which
what the verification code read, but way less than what you would
practically care about in most circumstances). If you have a weak
instruments GMM something with millions of observations, I won't be
surprised to see somewhat larger differences in results, may be closer
to c(epsfloat), but changes in p-values usually mean something
catastrophic happened, in terms of reproducibility. In Erik's shoes, I
would first make absolutely sure that I am using identical data sets
and identical code.
In the computer
a+(b+c) may differ from (a+b)+c
in the least significant bit. StataMP breaks up a problem into pieces and
that will affect the order in which arithmetic operations are conducted.
If the values are of greatly different magnitudes, that bit may make a
noticable difference in a sum or difference, but not if the problem is
well specified. If you are seeing differences in real data (as opposed to
malicious test datasets) that almost certainly signals statistical
problems, and not Stata bugs.
Daniel Feenberg
-- Stas Kolenikov, PhD, PStat (ASA, SSC)
-- Principal Survey Scientist, Abt SRBI
-- Opinions stated in this email are mine only, and do not reflect the
position of my employer
-- http://stas.kolenikov.name
On Tue, Jan 21, 2014 at 2:08 PM, Richard Williams
<[email protected]> wrote:
If true, that is a little scary. My hope and expectation would be that the main difference would be how quickly you get the answer, but that in the end the answer would be the same.
Sent from my iPad
On Jan 21, 2014, at 2:38 PM, David Kantor <[email protected]> wrote:
Hello,
I just want to add that I once had a situation in which I got different results in Stata MP vs. a single-processor version (of the same generation, I believe). I don't recall the details, but in the end, we decided to trust the Stata-MP result. Stata Technical Support explained that it had to do with MP making assumptions about what order some of the intermediary results could be computed -- or what could be concurrently. I could dig up the details of this matter, but at this point, the significant point is that, sometimes, MP can yield different results as compared to the single-processor version. Potentially, your two Statas could differ in that regard.
--David
At 05:54 PM 1/20/2014, David Radwin wrote in response to Erik Berwart:
It is difficult to tell why your results might differ from the limited information you provide here (see http://www.stata.com/support/faqs/resources/statalist-faq/#stata). Can you provide more details including sample data and code and actual output?
You also might use publicly available datasets to compare results from your installations with published accuracy test results. See http://www.stata.com/support/cert/ for details.
David
--
David Radwin, Senior Research Associate
Education and Workforce Development
RTI International
2150 Shattuck Ave. Suite 800, Berkeley, CA 94704
Phone: 510-665-8274
www.rti.org/education
-----Original Message-----
From: [email protected] [mailto:owner-
[email protected]] On Behalf Of ERIK BERWART ARAYA
Sent: Monday, January 20, 2014 1:03 PM
To: [email protected]
Subject: st: Can I have different results from Stata13 depending on the
OS?
Dear Statalisters,
I have installed Stata in my mac and in a virtual machine of windows8
Last week the regression was taking a lot of time on windows so I decide
to see if things were faster in the Mac version. I was shocked to see the
results were different (I mean not only the magnitude of the coefficients
but the lvalues, therefore, the interpretation using each version of
outputs were completely different).
Can anyone give me a hand here?
Regards,
Erik Berwart
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/faqs/resources/statalist-faq/
* http://www.ats.ucla.edu/stat/stata/
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/faqs/resources/statalist-faq/
* http://www.ats.ucla.edu/stat/stata/
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/faqs/resources/statalist-faq/
* http://www.ats.ucla.edu/stat/stata/
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/faqs/resources/statalist-faq/
* http://www.ats.ucla.edu/stat/stata/