Statalist


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

Re: st: Binomial regression


From   Roger Harbord <rogerharbord@bigfoot.com>
To   statalist@hsphsun2.harvard.edu
Subject   Re: st: Binomial regression
Date   Sat, 04 Aug 2007 12:14:09 +0100

Thanks Bobby, very interesting.  Personally I'm particularly interested
in your mention that this is the result of similar work for the log link
- presumably a log/logit link? Is this available somewhere, or will it be?

Thanks,
Roger.

Roberto G. Gutierrez, StataCorp LP wrote:
> Vince Wiggins and I responded on this thread yesterday, in which one of our
> main points was that forcing convergence when an identity link may not be
> appropriate results in a loss of model interpretation.  This is a case where
> not only is there no free lunch, but the exact price of that lunch is unclear.
>
> We mentioned various ways that convergence could be achieved.  One method I
> did not mention is the result of similar work I have done working with the log
> link function.  This method carries with it an ease of interpretation (akin to
> knowing the price of your lunch) and involves using a modified version of the
> identity link, something I call the "Identity/logit" link.
>
> <snip>


*
*   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 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index