Hi All,
At the recent Australian and New Zealand Users Group meeting in Sydney,
Bobby G stated that we can expect quite a bit more in the way of competing
risk methods in Stata (though, of course, no timelines were promised!).
Bobby did say that further functionality in this area might first appear in
the Stata Journal before any formal additions to Stata. I personally think
this area of activity would be a VERY GOOD THING; discussion and use of CR
methods seem to be increasing very rapidly, but major competitors like SAS
still have not got much to offer (or so say my SAS-o-phile colleagues).
Agree.
However, I would add that in respect to check the proportionality
assumption of the hazard of the subdistribution (the present question
asked by Eduardo) I believe that two options are already available
within -stccreg-:
1) It is possible to test the significance of a time-dependent effect of
a covariate by including its interaction with the time;
2) It is possible to estimate Schoenfeld residuals and then plot them
together with a smooth function of them.
This is not all we would hope, but it could be enough.
Enzo
Apart from the modelling domain, even some basic additions to the -sts list-
and -sts test- commands giving options to list the cumulative incidence
function (CIF) with confidence intervals, and perform Gray's and/or Pepe &
Mori's tests would be most welcome. I pay tribute to Enzo Coviello who has
provided some most useful programs in this area, but incorporation of these
and other facilities into official Stata would seem to be in order,
especially if Stata is tackling the modelling domain. Note in particular
that although one can extract the CIF from the results of the -stcrreg-
model, this is the predicted CIF, not the empirical distribution, as given
by Enzo's -stcompet-.
Of course, expenditure of StataCorp's resources in one direction presents a
competing risk for tackling their other priorities. But unlike the others, a
full competing risk suite would enable them to list, test and model their
problems :-)
Phil
Philip Ryan
Professor and Director
Data Management & Analysis Centre
School of Population Health & Clinical Practice
University of Adelaide
-----Original Message-----
From: [email protected]
[mailto:[email protected]] On Behalf Of Eduardo Nunez
Sent: Thursday, 26 November 2009 11:46 PM
To: [email protected]
Subject: st: incompleteness of stcrreg
Dear Statalisters,
I wonder why the new Stata command stcrreg did not come with
essentials postestimation commands, like "estat concordance" or ways
to test the proportionality assumption (which is a crucial requirement
for its use).
I tried to run "estat phtest" after stcrreg unsuccessful. I hope soon
they will be available. Switching to Cox in order to test the
proportionality assumption doesn't seems to be right.
Any thoughts?
Best wishes,
Eduardo
*
* 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/
No virus found in this incoming message.
Checked by AVG - www.avg.com
Version: 9.0.709 / Virus Database: 270.14.85/2531 - Release Date: 11/28/09
06:09:00
*
* 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/
--
Enzo Coviello
Epidemiology Unit ASL Bari
Piazza V. Emanuele, 14 - 70054 GIOVINAZZO (BA)
Italy
tel +39 080 3357883 - fax +39 080 3357867
mobile 347 5016016
^^^^
Home
via Imbriani, 186 - 70055 MINERVINO MURGE (BA)
tel +39 0883 695055
*
* 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/
