At 02:38 PM 4/26/2009, Clive Nicholas wrote:
(Slight change of subject, but...)
Richard Williams replied to Jay Verkuilen:
> Indeed, as the -intreg- documentation notes, -intreg- is a
generalization of
> the models fit by -cnreg- and -tobit-.
Really? Does this implicate -xtintreg- as well? I only ask because
I've recently used -xtintreg- to fit pooled TSCS models with an
ordinal-level dependent variable (the nearest best thing in Stata
9.2), and even then you have to assume random effects when using it
(unfortuately, there is no fixed-effect equivalent; but with my
football data, this condition wasn't too restrictive, and I could
always circumvent it by mean-centering). However, although my ODV
contained lots of 0s, it was not in any way censored.
I am not sure what you mean by "implicate" but I suppose that most or
all of what is true of intreg is also true for xtintreg. Also, I
wonder about your statement that the data are in no way censored. I
think the lower and upper bounds are supposed to be regarded as
negative and positive infinity, even if in practice the observed
range is much more limited than that.
FYI, I have a hypothetical example where intreg works spectacularly well:
http://www.nd.edu/~rwilliam/xsoc73994/intreg3.pdf
However, as noted in the handout, everything is set up so intreg's
assumptions are met. A simulation analysis examining how well it
worked when assumptions were violated would no doubt be useful.
-------------------------------------------
Richard Williams, Notre Dame Dept of Sociology
OFFICE: (574)631-6668, (574)631-6463
HOME: (574)289-5227
EMAIL: [email protected]
WWW: http://www.nd.edu/~rwilliam
*
* 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/
