Possibly isotonic regression, as estimated using the
'pool-adjacent-violators algorithm (PAVA), might be one solution. Wim
van Putten wrote a program -ira- to do isotonic regression analysis
using PAVA as part of a package -wvpreg- that used to be available
from his personal website, but unfortunately that no longer exists.
-ira- was written in Stata version 5.0 so appears a little
old-fashioned now but still works. The package gives Wim van Putten's
email address as [email protected] -- a quick check on Google
Scholar suggests he's still at Erasmus University Medical Center, so
anyone wishing to obtain the package could try emailing him.
Roger.
--
Roger Harbord
http://www.epi.bris.ac.uk/staff/rharbord.htm
On Wed, Jan 20, 2010 at 8:49 PM, <[email protected]> wrote:
> Would not the poster's particular problem be solved by a version of
> the "pool-adjacent-violators" algorithm?. See:
> http://fedc.wiwi.hu-berlin.de/xplore/ebooks/html/anr/anrhtmlnode43.html
>
> Steve
>
> On Wed, Jan 20, 2010 at 3:13 PM, Maarten buis <[email protected]> wrote:
>> --- On Wed, 20/1/10, Dana Chandler wrote:
>>> Does anyone else have a possible
>>> answer? I don't think the below
>>> response has answered this question.
>>
>> You may not like the answer, but it did
>> answer your questions. The problem is that
>> unless someone has already implemented that
>> in a program and submitted it to SSC, any
>> respons will be similar to the Stata FAQ
>> you already refered to. You can't expect an
>> answer that is more elaborate (or even as
>> elaborate) as that FAQ.
>>
>> -- Maarten
>>
>>
>>>
>>> Although I appreciate the book reference and confirmation
>>> that the
>>> article I mentioned is applicable to oprobit or any other
>>> model, the
>>> below post has not provided any guidance on how to
>>> proceed.
>>>
>>> Have any other researchers solved this specific problem,
>>> since I don't
>>> think it is that unusual of a problem.
>>>
>>> Thanks in advance,
>>> Dana
>>>
>>>
>>>
>>> On Wed, Jan 20, 2010 at 2:36 AM, Maarten buis <[email protected]>
>>> wrote:
>>> > --- On Tue, 19/1/10, Dana Chandler wrote:
>>> >> oprobit y pop_1 pop_2 pop_3 pop_4
>>> >>
>>> >> ... and I would like to constrain each successive
>>> >> population parameter has a coefficient lower than
>>> the
>>> >> previous so that (pop_2-pop_1)>=0,
>>> (pop_3-pop_2)>=0, etc.
>>> >>
>>> >>
>>> >> I recognize that there has been a thread (and faq)
>>> from a
>>> >> few years back (<http://www.stata.com/support/faqs/stat/
>>> > intconst.html>) explaining how to set up interval
>>> (non-linear)
>>> >> constraints using ML to perform a linear
>>> regression. However,
>>> >> even though the article suggests that I could use
>>> the similar
>>> >> methodology to derive it for probits, I'm not 100%
>>> sure that
>>> >> it's directly applicable in my case and for
>>> oprobits.
>>> >
>>> > It is also applicable to oprobit, or any other model.
>>> >
>>> >> I have not worked with stata's ML function before.
>>> Does
>>> >> anyone have any advice?
>>> >
>>> > If you are serious about getting into this type of
>>> modeling then
>>> > you can't go wrong by getting: William Gould, Jeffrey
>>> Pitblado,
>>> > William Sribney (2006) Maximum Likelihood Estimation
>>> with Stata.
>>> > College Station: Stata Press.
>>> > http://www.stata.com/bookstore/mle.html
>>> >
>>> > Hope this helps,
>>> > Maarten
>>> >
*
* 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/
