Re: st: Problems with eststo, stata versioning

[Nick Sanders <sandersn@stanford.edu>]

Hi Ryan,

Have you recently updated your Stata? I believe some commands changed how they calculate standard errors (e.g., areg) between versions, which might explain the standard error difference. Older versions of Stata used to "remove" omitted variables in regression results, but in some commands I believe Stata 11 keeps them (with no output), which is what made me suspicious it may be update related.

Best,
A different Nick

On Jan 3, 2012, at 12:51 PM, Nick Cox <njcoxstata@gmail.com> wrote:

> Before anyone has a chance to go "tu quoque" I will underline that
> -eststo- is from SSC and part of the -estout- package.
> 
> Ryan will find that changes to that package are carefully documented at
> 
> http://repec.org/bocode/e/estout/history.html
> 
> Despite not being a betting person, I would bet that -eststo- is not
> in any sense changing standard errors and that your main problem lies
> elsewhere.
> 
> Nick
> 
> On Tue, Jan 3, 2012 at 5:55 PM, Nick Cox <njcoxstata@gmail.com> wrote:
> 
>> -eststo- is a user-written program. Updating or upgrading your installation
>> of Stata makes absolutely no difference to any installations of user-written
>> programs, which remain entirely your own responsibility. However, Stata does
>> provide -adoupdate- as a convenience command to help in maintaining your
>> collection of user-written add-ons.
> 
> On 3 Jan 2012, at 16:44, Ryan Turner <rjturner@cmu.edu> wrote:
> 
>>> I am having a problem with the -eststo- package.  Although I believed my
>>> Stata 11.2 installation to be up-to-date, I found that I have been using an
>>> older version of -eststo-, from May 2007.  It appears that Stata did not
>>> recognize the newer version (from October 2009) as being an update to the
>>> same package.  I discovered this when searching for -estpost-, which did not
>>> exist in the 2007 version.  I decided to force update the package, which
>>> caused some of my thesis regression results to change in significance
>>> (coefficients the same, but standard errors changed).  In addition,
>>> variables dropped for collinearity are retained in regression output as:
>>> 
>>> dropped_var   0     0     0     0
>>>            (.)   (.)   (.)   (.)
>>> 
>>> Why would standard errors change when processing regression output through
>>> -eststo- and -esttab-?  Should I consider former output as incorrect,
>>> perhaps because of an old bug in -eststo-, and that new output is correct?
>>>  Is there some option I may have set that is doing strange things to my
>>> results?  Is it a new feature to retain omitted variables as zero with
>>> missing standard errors?  Why did Stata not recognize the package as an
>>> update?
>>> 
> 
> *
> *   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/

*
*   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/