Geroff, The answer to your question is that the answer depends on the question that you wish to ask. If you want to know if men have a higher probability of mortality (or mortality rate) than women after taking into account differential alcohol use, red meat consumption, etc. by sex then you would adjust for these factors. If your question is do men have a higher probability of mortality than women then you would not adjust for the various potential confounders. Adjusting for confounders can be very important. Consider a study of coffee drinking as a factor influencing mortality. Such a study may well find that coffee drinking is related to mortality, even if coffee drinking is completely innocuous. Why might this be so? Smoking is associated with mortality, and smoking is associated with coffee drinking (some people smoke while the drink coffee). If you fail to adust for smoking, you may be lead to an incorrect inference about the relation between coffee and mortality. I hope this helps. John John Sorkin M.D., Ph.D. Chief, Biostatistics and Informatics Baltimore VA Medical Center GRECC, University of Maryland School of Medicine Claude D. Pepper OAIC, University of Maryland Clinical Nutrition Research Unit, and Baltimore VA Center Stroke of Excellence
University of Maryland School of Medicine Division of Gerontology Baltimore VA Medical Center 10 North Greene Street GRECC (BT/18/GR) Baltimore, MD 21201-1524 (Phone) 410-605-7119 (Fax) 410-605-7913 (Please call phone number above prior to faxing) [EMAIL PROTECTED] >>> "Geoff Russell" <[EMAIL PROTECTED]> 9/20/2006 9:54 AM >>> Peter et al, Thanks for the reply, I did reread the posting guide before posting and figured it was a short question and might just have a short answer. I have Therneau's book on order, which will probably clarify the matter in time. I understand stratifying to deal with confounding, but not adding it as a covariate in a regression. e.g, If one of the gender related effects you mention happens to be drinking, then we don't want to "get rid of it", it may well be an additional covariate and we want its full effect embodied in the b value for that covariate. I'll keep reading! Cheers, Geoff On 20 Sep 2006 14:47:00 +0200, Peter Dalgaard <[EMAIL PROTECTED]> wrote: > "Geoff Russell" <[EMAIL PROTECTED]> writes: > > > Hi useRs, > > > > Many studies of the link between red meat and colorectal cancer use > > Cox proportional > > hazards with (among other things) a gender covariate. > > > > If it is true that men eat more red meat, drink more alcohol and smoke more > > than > > women, and if it is also true that alcohol and tobacco are known risk > > factors then why does > > it make sense to "adjust" for gender? I would think that in this > > case some of the > > risk that should be properly attributed to the bad habits will actually end > > up being attributed to being male instead. > > This is more than a bit off-topic for the list, but in (very) brief: > Because you need to get rid of purely gender related effects that > disturb the analysis and may create spurious association. > > Otherwise you would become able to "prove" effects like stiletto heels > causing breast cancer, etc. > > -- > O__ ---- Peter Dalgaard Øster Farimagsgade 5, Entr.B > c/ /'_ --- Dept. of Biostatistics PO Box 2099, 1014 Cph. K > (*) \(*) -- University of Copenhagen Denmark Ph: (+45) 35327918 > ~~~~~~~~~~ - ([EMAIL PROTECTED]) FAX: (+45) 35327907 > ______________________________________________ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R ( http://www.r/ )-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code. Confidentiality Statement: This email message, including any attachments, is\ for the s...{{dropped}}
______________________________________________ R-help@stat.math.ethz.ch mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
