Re: st: Incompatibility of base level operators with vce(jackknife)

[jpitblado@stata.com (Jeff Pitblado, StataCorp LP)]

David Muller <davidmull@gmail.com> found a parsing conflict between
-vce(jackknife)- (and -vce(bootstrap)-) and factor variables varlist notation
that uses parentheses:

> Factor variable base level operators such as -ib(freq).- seem to be
> incompatible with the option -vce(jackknife)-:
> 
> .sysuse auto
> .reg mpg ib(freq).foreign, vce(jackknife)
> variable ib not found
> r(111);
> 
> This is odd, since specifying jackknife variance estimation with the
> prefix -jackknife:- does not trigger an error.
> 
> . jackknife: reg mpg ib(freq).foreign
> (running regress on estimation sample)
> -- output omitted --
> 
> Both Stata 11.2 and 12 exhibit this behavior. Is this intended? It
> doesn't seem to be documented in [U] Factor variables, or [R]
> vce_option. Obviously it is easy to work around the problem by
> generating indicator variables manually, but I can't think why full
> factor variable notation should not be allowed when specifying
> -vce(jackknife)-.

David found a bug in the parsing code that is called when -vce(jackknife)- or
-vce(bootstrap)- options are specified.  We have determined the source of the
parsing problem and will have it fixed in a future Stata 12 ado-file update.

In the mean time, David can bind the varlist in parentheses to prevent the
error:

	reg ( mpg ib(freq).foreign ) , vce(jackknife)

--Jeff
jpitblado@stata.com
*
*   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/