Re: Re: st: Different Stata versions give different standard errors

[Kelly Jones <kmjones@gmail.com>]

Thanks Richard and Nick.

It seems the problem is indeed with the "areg" command in version 11
(and 10 I guess?).  The "xtreg, fe" estimation reproduces the standard
errors of the "areg" estimation under version 9.  Oddly, prefixing the
"areg" command in version 11 with "version 9:" does not fix the
problem.

This is technically an error with the "areg" estimation in the newer
versions, then?  Is Stata Corp aware of this?  Your explanation of it
makes sense for why my SE would be so different -- since this
estimation has many fixed effects that drop out.

Many thanks for your helpful replies.

Kelly



Date: Sun, 7 Aug 2011 16:53:58 -0700
From: Nick Sanders <sandersn@stanford.edu>
Subject: Re: st: Different Stata versions give different standard errors

Though I can't provide exact details (but would love to hear from
someone who can), I believe there was a change to the degrees of
freedom calculation between Stata 9 and 10 for areg, where the
(omitted) FEs were ignored in 9 but now counted as coefficients to be
estimated in 10. You will find xtreg, fe to be consistent across
versions (I think), and agreeing with the areg from version 9.

On Aug 7, 2011, at 5:30 PM, Richard Williams
<richardwilliams.ndu@gmail.com> wrote:

> At 05:05 PM 8/7/2011, Kelly Jones wrote:
>> Hello all, Im having the following issue and wondering if anyone has
>> any experience of this or any suggestions of its cause:
>>
>> Running the following command on the exact same data set, I get very
>> different standard errors when using different versions of Stata.  The
>> command is:
>>
>> areg var1 … var9 if var10 & var11 & var12 & var13==1 [pw=var14],
>> absorb(var15) cluster(var16)
>>
>> When using Stata SE 9.2 for Macintosh (on Mac OS 10.6  machine has
>> Intel core 2 duo processor, but Stata version is not for multi-core).
>> the SE on the coeff for var1 is .019;
>>
>> When using Stata MP 11.2 on Windows 7 (machine has Intel core i7
>> processor) the coefficient estimate is the same but the SE is .032!
>
> On the win7 machine, I would try prefixing your command with version 9.2:
>
> Then, if the standard error is .019, it will suggest the change is due to some change in areg across versions of Stata. If it is .032, it will suggest something about differences across hardware platforms.
>
>
> -------------------------------------------
> Richard Williams, Notre Dame Dept of Sociology
> OFFICE:             (574)631-6668      ,             (574)631-6463
> HOME:               (574)289-5227
> EMAIL:  Richard.A.Williams.5@ND.Edu
> WWW:    http://www.nd.edu/~rwilliam
>

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/