[R] stepwise selection cox model

[David Winsemius]

On May 25, 2011, at 12:11 PM, linda Porz wrote:

> Many thanks for your reply. I have run a stepwise selection in Stata  
> and R using the function fastbw (rule="p") from Design package. Both  
> functions give the same results. Is this because both functions do  
> the same job or can it be that for different data one will have  
> different results?

I don't understand your question. Why would "giving the same results"  
be a concern? And why would one expect that with different data one  
would _not_ get different results? The point of the critique against  
stepwise procedures is that they assume too much "determinism" (i.e.  
that all of the internal structure of the small sample of data will be  
present in the wider universe) and that they generate too much  
"confidence" on the part of the unwary and insufficiently educated user.

-- 
David.

>
> Many thanks,
> Linda
>
>
>
> 2011/5/25 Bert Gunter <gunter.berton at gene.com>
> See the Vignette in the glmnet package for one alternative approach to
> variable selection. Of course, you need to gain some background to
> know what you're doing here.
>
> -- Bert
>
> On Wed, May 25, 2011 at 8:38 AM, Marc Schwartz  
> <marc_schwartz at me.com> wrote:
> > Hi,
> >
> > You are unlikely to find one, as fundamentally, stepwise  
> procedures are a bad way to engage in covariate selection. Search  
> the list archives at rseek.org using 'stepwise' as the keyword to  
> see a plethora of discussion on this point.
> >
> > This is not a new issue BTW, as I happened to stumble upon this  
> 1998 Stata FAQ recently during a related search:
> >
> >  http://www.stata.com/support/faqs/stat/stepwise.html
> >
> > and there are more recent literature citations and books that  
> reinforce those points.
> >
> > HTH,
> >
> > Marc Schwartz
> >
> > On May 25, 2011, at 4:28 AM, linda Porz wrote:
> >
> >> Sorry, I have wrote a wrong subject in the first email!
> >>
> >> Regards,
> >> Linda
> >>
> >> ---------- Forwarded message ----------
> >> From: linda Porz <linda.porz at gmail.com>
> >> Date: 2011/5/25
> >> Subject: combined odds ratio
> >> To: r-help at r-project.org
> >> Cc: r-help-request at stat.math.ethz.ch
> >>
> >>
> >> Dear all,
> >>
> >> I am looking for an R function which does stepwise selection cox  
> model in r
> >> (delta chisq likelihood ratio test) similar to the stepwise, pe  
> (0.05) lr:
> >> stcox in STATA.
> >>
> >> I am very thankful for any reply.
> >>
> >> Regards,
> >> Linda
> >
> > ______________________________________________
> > R-help at r-project.org 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.
> >
>
>
>
> --
> "Men by nature long to get on to the ultimate truths, and will often
> be impatient with elementary studies or fight shy of them. If it were
> possible to reach the ultimate truths without the elementary studies
> usually prefixed to them, these would not be preparatory studies but
> superfluous diversions."
>
> -- Maimonides (1135-1204)
>
> Bert Gunter
> Genentech Nonclinical Biostatistics
>

David Winsemius, MD
West Hartford, CT