R help - Jan 2014 - Type III tests and Cox models

Someone asked a question about this on the list a couple months ago. I replied
that I
didn't know any sound basis for it, and the counter-reply was "SAS has
it".  Grant
applications were due and I got testy, so there was no further progress at that
point.
   As the author of the survival package I really did want to understand the
issue
however.  I've now done my homework, and the result is contained in a 30+
page vignette
that has been added to the most recent release of the package.  The short
summary: I was
suspicious before, now I know for certain that it is misguided, and the phreg 
implementation of the idea is worse.
   If you are interested take a look.  If significant debate arises about the
statistical
issues therein I don't think R-help is quite the right forum for discussion,
and will
entertain any idea wrt a more appropriate venue.  Rehashing long and old
standing
controversies about type 3 for linear models is, however, not exciting to me.

Terry Therneau

On 20 Jan 2014, at 18:47 , Terry Therneau <therneau at mayo.edu> wrote:
>  The short summary: I was suspicious before, now I know for certain that it
is misguided, and the phreg implementation of the idea is worse.
A fortune candidate, if ever I saw one.

-pd

-- 
Peter Dalgaard, Professor,
Center for Statistics, Copenhagen Business School
Solbjerg Plads 3, 2000 Frederiksberg, Denmark
Phone: (+45)38153501
Email: pd.mes at cbs.dk  Priv: PDalgd at gmail.com

To save others a little hunting, you can read the vignette online at
http://cran.r-project.org/web/packages/survival/vignettes/tests.pdf

Hadley

On Mon, Jan 20, 2014 at 11:47 AM, Terry Therneau <therneau at mayo.edu>
wrote:
>  Someone asked a question about this on the list a couple months ago. I
> replied that I didn't know any sound basis for it, and the
counter-reply was
> "SAS has it".  Grant applications were due and I got testy, so
there was no
> further progress at that point.
>   As the author of the survival package I really did want to understand the
> issue however.  I've now done my homework, and the result is contained
in a
> 30+ page vignette that has been added to the most recent release of the
> package.  The short summary: I was suspicious before, now I know for
certain
> that it is misguided, and the phreg implementation of the idea is worse.
>   If you are interested take a look.  If significant debate arises about
the
> statistical issues therein I don't think R-help is quite the right
forum for
> discussion, and will entertain any idea wrt a more appropriate venue.
> Rehashing long and old standing controversies about type 3 for linear
models
> is, however, not exciting to me.
>
> Terry Therneau
>
> ______________________________________________
> 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.


-- 
http://had.co.nz/