Thank you very much for all the comments and suggestions.
I haven't come across any literature talking about "Generalised interval
regression". I just used that term because I needed to highlight the
similarity in what I needed to the generalised ordered probit and logit
models, I am sorry if I created any confusion.
I have already used the -intreg- and I have tried some of the
suggestions, including taking the logs of the intervals.
The use of logs in the intervals imply a change in the effect of the
explanatory variables when changing intervals. However, initially I
thought that I should ask in the statalist if anybody has attempetd
something like that before, because I know the originality of the people
in the list.
Since it is not an easy idea to implement and I know I am not able to
program or develop it, I think I ll leave it for the moment and stick
with -intreg- and goprobit.
Thank you
Manos
-----Original Message-----
From: [email protected]
[mailto:[email protected]] On Behalf Of Richard
Williams
Sent: 04 December 2006 18:09
To: [email protected]
Subject: RE: st: Generalised interval regression
At 10:14 AM 12/4/2006, Maarten Buis wrote
>With such a dependent variable, I would look at -intreg-. If you want
>to think that the effects for the different intervals differ, then you
>think that the effect isn't linear in the latent variable (hours of
>activity per week). If you specify the functional form you expect that
>effect to have, you could make some progress along those lines, i.e.
>build your own likelihood function. However, I would at least start
>with just -intreg- and the assumption of a linear effect on hours of
activity.
Now that I think about it, let me second the suggestion of Maarten and
others, and add a recommendation that you read the reference manual
entry on intreg. That entry shows (a) how to test the appropriateness
of the intreg model, e.g. you can do a contrast between oprobit and
intreg, and (b) if intreg is problematic, it suggests transforming the
dv, e.g. take the logs of the interval boundaries. If the goal is "to
give different effects for explanatory variable according to the levels
in the dependent" then taking logs or doing some other transformation of
Y may accomplish this goal.
Again, I've never heard of a "generalised interval regression model"
but if there is a literature on it I'd like to see it and might even be
willing to take a crack at programming it. But, I'm not smart enough
just to do it on my own! In the absence of a clear idea on how to
proceed with your original idea, I would either work on salvaging intreg
or just going with an oprobit or goprobit model.
-------------------------------------------
Richard Williams, Notre Dame Dept of Sociology
OFFICE: (574)631-6668, (574)631-6463
FAX: (574)288-4373
HOME: (574)289-5227
EMAIL: [email protected]
WWW (personal): http://www.nd.edu/~rwilliam
WWW (department): http://www.nd.edu/~soc
*
* 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/
*
* 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/
