Re: st: Testing equality of coefficients in different logistic models

[Richard Williams <Richard.A.Williams.5@ND.Edu>]

At 11:38 AM 9/13/2006, Sara Mottram wrote:
> I would also value opinions on whether there is any value in 
> performing two binary regressions, given that -gologit2- has an 
> automatic fitting procedure in -autofit-. (I am basing my analysis 
> plan on a fairly out of date paper, which suggests that binary 
> regressions are the only way to choose where non-proportionality is 
> needed as there were no programs able to do this at the time).
I would suggest downloading and installing Long and Freese's 
-spost9-, which includes the -brant- command.  This will basically 
automate what you propose doing.

My experience is that these different procedures tend to lead you to 
the same conclusions, but not always.

Whether or not you use autofit may depend on whether or not you have 
any theory as to where the non-proportionalities are.  Sometimes 
people do, but often they don't.  I always say that autofit is 
perhaps the lesser of 3 evils.  You don't like to use ologit if its 
assumptions are violated; but if you switch to mlogit you may be 
estimating far more parameters than is necessary.  autofit lets you 
hit a middle-ground, but you should keep in mind that it relies on 
mindless empiricism to get you there!

For more on gologit2, including documentation on its use and some 
relatively recent stuff on the rationale behind it, see

http://www.nd.edu/~rwilliam/gologit2/index.html



-------------------------------------------
Richard Williams, Notre Dame Dept of Sociology
OFFICE: (574)631-6668, (574)631-6463
FAX:    (574)288-4373
HOME:   (574)289-5227
EMAIL:  Richard.A.Williams.5@ND.Edu
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/