Richard wrote:
> One thing that occurs to me is
> that users
> might want to write their publicly shared -ado- files with -estadd-
> compatibility in mind, i.e. return results in a format that
> would make it
> easy to write an -estadd- routine for. For example (and
> definitely not the
> best example!) I could probably rewrite my -pcorr2- routine
> to return the
> partial and semipartial correlation coefficients in r() in
> addition to just
> printing them out.
I think this is a very general point: Any statistical routine should
return its results in r() or in e().
> I haven't looked carefully through your documentation, but it
> if doesn't
> exist already perhaps you could have some instructions on how
> a program
> needs to return results to make it easily compatible, e.g.
> how do the rows
> and columns need to be named? I imagine -vif- would have
> been easier to
> add if it returned its results in a matrix rather than as a bunch of
> scalars and macros.
It does not really matter how exactly a program returns its results
as long as it returns them at all. The vif command, for example,
uses r()-scalars for the vif-values and r()-macros for the variable
names. This is perfectly fine. However, a matrix would be okay, too.
In case of -pcorr2- I would prefer using matrices (row vectors, to be
precise), since two statistics (partial correlations and semipartial
correlations) are to be returned.
ben
*
* 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/