Michael Rosenfeld responded to a posting from Veit Grote, who is having
difficulty with -desmat- under Stata Release 8:
--------------------------------------------------------------------------------
On July 16, the following message was posted to Statalist:
. . .
"desmat" under Stata8 does not work as under Stata7. The nice feature of
displaying labels of categorical variables disappeared somehow. Maybe
someone has a clue how to fix it?!
Thanks
Veit Grote
As far as I can tell, Mr. Grote's question has not previously been
answered. I initially ran into the same problem with desmat and Stata
8. The problem was that I was using an out-of-date version of -desmat-.
. . .
version 3.1 and after of -desmat- appear to work fine with Stata 8. I run
Stata 8.1 on Windows 2K and Windows 98 machines.
I should add that -desmat- is a very helpful program that is capable of
building higher order interactions than -xi- or -xi3-, and -desmat- also
produces much nicer looking output.
. . .
--------------------------------------------------------------------------------
I don't know about displaying value labels with -desmat-, never having paid
attention to value labels in the past with -desmat-, but I also couldn't see
them in the printout, just as Veit mentioned, even with the latest update
(2003-09-25) of -desmat- from SSC.
There is another a minor glitch with -desmat- and Stata Release 8.X, but I've
only noticed it with -xtgee-. The attached do-file shows where it will crop up
and how to work around it. The workaround (include -nomodinfo- in the option
-desrep()-) is so easy that I haven't bothered to contact the author about bug.
-desmat- appearently has trouble finding the model P-value associated with the
overall Wald chi-square test for -xtgee-.
Joseph Coveney
P.S. I agree that -desmat- is a fabulous command. It's become the default for
production work in any situation where interactions are evaluated or where
there are more than two categories to a predictor. It's been a godsend where
the categories of a predictor have a natural ordering. Congratulations and
appreciation go to John Hendrickx for adding a very valuable capability to
Stata and for continuing to support it.
--------------------------------------------------------------------------------
clear
set more off
set seed 20030927
set obs 210
generate int pid = _n
generate byte dos = mod(_n, 3) + 1
generate float bas = invnorm(uniform())
forvalues t = 1/3 {
generate float res`t' = bas * sqrt(0.7) + ///
(1 - 0.7) * invnorm(uniform())
quietly replace res`t' = res`t' * 0.5 * dos * (`t' - 1) / 2
}
reshape long res, i(pid) j(tim)
char tim[pzat] orp(2)
char bas[pzat] dir
char dos[pzat] hel(b)
// No problem with the following line . . .
desmat: xtreg res dos*tim*bas, verbose desrep(all) i(pid) fe
// . . . but the following line will produce an error.
* desmat: xtgee res dos*tim*bas, desrep(all) i(pid)
// The following line works, however.
desmat: xtgee res dos*tim*bas, verbose desrep(nomodinfo all) i(pid)
exit
--------------------------------------------------------------------------------
*
* 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/
