Re: st: exchanging the outcome and exposure in logistic reg

[Steven Samuels <sjhsamuels@earthlink.net>]

--
Ricardo:

It is correct to do the prospective logistic analysis for unmatched  
case-control studies. See: Prentice, R. L., & Pyke, R. (1979).  
Logistic disease incidence models and case-control studies.  
Biometrika, 66(3), 403-411;  Carroll et al., 1995, Journal of the  
American Statistical Association, 90, 157-169.  For matched studies,  
the appropriate analysis is conditional logistic regression of the  
case-status on covariates (-clogit-).  See: Fleiss, J. L., Levin, B.  
A., & Paik, M. C. (2003). Statistical methods for rates and  
proportions (3rd ed ed). Hoboken, N.J: Wiley, 411-416.

-Steve

On Jan 24, 2009, at 1:47 PM, Ricardo Ovaldia wrote:
> I recall a discussion on the list a while back regarding this  
> topic, but I did find it in the archives.
>
> In a case-control study the traditional way of modeling a logistic  
> regression is:
>
> logistic case died x1 x2 x3 x4
>
> where case is the presence of cancer (0/1) for example.
>
> Now, an investigator wants me to fit:
>
> logistic died case x1 x2 x3 x4
>
> and I am not sure that this is correct any longer. I recall that it  
> is okay to “flip” the outcome and exposure variables if there are  
> no additional covariates, however, with covariates (x1, x2 …) that  
> is not okay anymore. Is that correct?

*
*   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/