Yes indeed. How stupid of me not to see that. Thanks
very much.
Nick
[email protected]
Jean Marie Linhart, StataCorp LP
> Al Feiveson asked:
>
> > Nick I was using -storecmd- to do -xtintreg- but I obtained slightly
> > different results after the -storecmd- statement and then if I
> > repeat the analysis with a straight command. For some reason, with
> > "storecmd", the integration method used is "ghermite" (not the
> > default), whereas with the straight command, it correctly uses the
> > default "aghermite". But I never issued an option for "ghermite" so
> > what's going on?
>
> Nick Cox replied:
>
> > I can't explain this, but I don't see that it's anything to do with
> > -storecmd-. -storecmd- doesn't reach inside your command and change
> > it. Nor can I think of side-effects that would cause this. Still,
> > lack of imagination is not proof of absence.
>
> I can explain what is going on. -storecmd- sets -version 6.0- in its
> code, so when it runs the -xtintreg- command, it is run as version 6.
> Under version control, -xtintreg- properly reverts to the old
> method of
> Gauss-Hermite quadrature which was used in versions prior to 9.
>
> When Al Feiveson runs his code from the command line, he is
> using version
> 9, and he gets version 9 behavior.
>
> We (at StataCorp) run into this sort of issue with version control
> frequently (crack the hood on -xtintreg- itself for
> example...) if Nick
> wants to change the behavior of -storecmd-, I'd suggest he change the
> portion of his code that executes the command from:
>
> * execute command
> `cmd'
>
> to:
>
> * execute command
> local vv : di "version " string(_caller()) ", missing :"
> `vv' `cmd'
>
> Hope that helps.
>
> --Jean Marie
> [email protected]
*
* 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/
