Re: st: error: variable present in model more than once when using

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

Brent Fulton <fultonb@berkeley.edu> wrote in that -svy jackknife- was
reporting the following error message:

	... present in model more than once
	r(322);

where '...' was the name of a predictor variable in the model.

This particular error message is only possible when there are multiple long-
and similarly-named predictors in the model being fit by -svy jackknife-.

The problem here is that -svy jackknife- (and other prefix commands such as
-svy brr-, -jackknife-, and -bootstrap-) try to use the original equation and
variable names when the elements of e(b) are being replicated.  For example,
suppose we have

	. svy jackknife: nbreg y x1 x2 x3

-svy jackknife- will compose variable names for the -keep- and -saving()-
options.  In this case the name stripes and corresponding potential variable
names are:

	name stripe		-> varname
	-------------------------------------------
	y:x1			-> y_b_x1
	y:x2			-> y_b_x2
	y:x3			-> y_b_x3
	y:_cons			-> y_b_cons
	lnalpha:_cons		-> lnalpha_b_cons

When the name of a potential variable exceeds 32 characters (see c(namelen) in
-help creturn-) then an abbreviated variable name is attempted by taking the
first 12 characters of the equation name and similarly for the variable name.

Brent has 2 predictor variables with around 30 characters in their name, but
they only differ in that last 9 characters and the response variable name is
10 characters long.

For the time being, Brent can solve this name conflict by shortening some of
his variable names.

We will take a serious look at improving the name generation logic used by
-svy jackknife- for the case of long- and similarly-named predictors.

--Jeff
jpitblado@stata.com
*
*   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/