Statalist The Stata Listserver


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

RE: st: Generalised interval regression


From   "Feiveson, Alan H. \(JSC-SK311\)" <[email protected]>
To   <[email protected]>
Subject   RE: st: Generalised interval regression
Date   Mon, 4 Dec 2006 10:17:19 -0600

One possible problem with using -intreg- is that it (-intreg-) assumes a
normally distributed latent variable that you can only observe in terms
of intervals. On the surface, it appears that hours of activity would
not be normally distributed for fixed values of the covariate(s). A
user-written ml-procedure with a nonnegative latent variable
distribution such as Gamma, that allows interval censoring might be more
appropriate. Perhaps it wouldn't be that difficult to modify intreg.ado
for this purpose.

Al Feiveson

-----Original Message-----
From: [email protected]
[mailto:[email protected]] On Behalf Of Maarten Buis
Sent: Monday, December 04, 2006 9:14 AM
To: [email protected]
Subject: RE: st: Generalised interval regression

--- Mentzakis, Emmanouil wrote: 
> I have a dependent variable (hours of an activity per week) that is 
> given in intrevals. Thus, 0-20, 21-60, 61-100.
> 
> However, if I understand correctly, the goprobit and the gologit2 are 
> able to give different effects for explanatory variable according to 
> the levels in the dependent (i.e relax the parallel line assumption).
> 
> And I was wondering if there is any code to do the same thing when the

> dependent varaiabel is not specified simply as ordinal but as
intervals.
> 
> I know that I could recode my dependent in order to have it as ordinal

> (i.e. no intervals specified) but I would like to use the intrevals 
> information that I have (i.e. the cut-offs would not need to be 
> estimated). Additionaly the knowledge of the cut-offs allows me to 
> know the scale of the variable.

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. 

HTH,
Maarten

-----------------------------------------
Maarten L. Buis
Department of Social Research Methodology Vrije Universiteit Amsterdam
Boelelaan 1081
1081 HV Amsterdam
The Netherlands

visiting address:
Buitenveldertselaan 3 (Metropolitan), room Z434 

+31 20 5986715

http://home.fsw.vu.nl/m.buis/
-----------------------------------------



*
*   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/



© Copyright 1996–2024 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index