Re: st: Individual risks scores using competing risks

[Steve Samuels <sjsamuels@gmail.com>]

Lucy Riley (lucy.riley89@googlemail.com):

Your original problem is:


> "CIF(t|X) = 1 - exp( - exp(X * beta^t. H0(t) ) where H0(t) is the cumulative sub-distribution hazard for the event at time t....My main problem is that I haven't come across any methods for finding H0(t)."

You responded to Ariel:

> Again the command "predict basecif" does not help either as this gives baseline CIF at time 0, which for a person with
> zeroed coefficients is set to be 1.

Your statement is incorrect.

Both -stcurve- and -predict- after -stcrreg- produce baseline CIFs at
non-zero times: the observation times of the individuals in the
estimation sample. (An incidental error is that CIFs = 0 at t = 0.)


The relation of a cumulative hazard to a CIF is

H = -log(1-CIF)

which you could have found by setting X to 0 in your equation. Note that
it's not necessary to work with H , as:

CIF(t|x) = 1 - (1 -CIF(t|0))^exp(xbeta)

You can get x'beta from:


. predict xscore , xbeta

.stcurve, cif outfile(myfile)

Then -cross- the outfiled data set (it will contain _t and cif1) with
your original data and you will have CIFs for every individual in the
estimation sample at every time in the estimation sample.

What do you intend to do with these values? I suggest that you reread
the "The case for competing risks regression" in the Remarks and
examples" section of the Stata 13 manual. See also:
http://www.stata.com/statalist/archive/2013-09/msg00089.html and related
messages about the places of -stcrreg- and -stcox- in a competing risks
analysis.

Steve


> 
> On Sep 13, 2013, at 10:06 AM, Lucy Riley wrote:
> 
> 
> I have looked at -stcompet- but this doesn't give me what I need.
> This program generates the Cumulative Incidence values for each person within the data set,
> this isn't helpful for my problem as I am trying to develop a risk score that can be used elsewhere.
> 
> What I really need is the baseline cumulative sub-distribution hazard at time 5years so that I can write down the
> formula for an individuals CIF give that persons age, sex etc.
> 
> Again the command "predict basecif" does not help either as this gives baseline CIF at time 0, which for a person with
> zeroed coefficients is set to be 1.
> 
> But still, Thank you very much for your help!
> 
> -
> -Date: Thu, 12 Sep 2013 11:37:33 -0400
> -From: "Ariel Linden, DrPH"<ariel.linden@gmail.com>
> -Subject: re: st: Individual risks scores using competing risks
> -
> -Have you looked at - stcompet -  (http://fmwww.bc.edu/RePEc/bocode/s/), or
> -in Stata v13 (perhaps earlier versions as well) using "predict , basecif"
> -after -stcrreg- ?
> -
> 
> 

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/


*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/