Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: SE's on survival estimates from parametric models


From   Maarten buis <maartenbuis@yahoo.co.uk>
To   stata list <statalist@hsphsun2.harvard.edu>
Subject   Re: st: SE's on survival estimates from parametric models
Date   Mon, 27 Apr 2009 09:16:42 +0000 (GMT)

--- On Sat, 25/4/09, Mark Hebblewhite wrote:
> Thanks for your suggestion, it worked! But it produced, as
> expected with the predict command, a prediction of the 
> mean survival rate and SE for each individual row/animal 
> in my dataset.
> 
> To obtain one estimate of survival for say, the entire
> dataset, would it then be appropriate to just use summary 
> statistics to calculate group-level means and SE's? 

Below is an example of how to compute the hazards, se, and 
confidence intervals at t=15 for someone of average age, 
but still differentiated by drug. If you want one number,
then also replace the variable drug with one representative 
value e.g. the mode.

*------------- begin example -------------------
sysuse cancer, clear
stset studytime, failure(died)
xi: streg i.drug age, dist(weibull)

preserve
local t = 15
local p "exp([ln_p]_b[_cons])"
local h0 "`p' *`t' ^(`p' -1)"

sum age if e(sample), meanonly
replace age = r(mean)
bys drug : keep if _n == 1 

predictnl hazard = `h0' * exp(xb()), ci(lb ub) se(se)
list drug hazard se lb ub, clean noobs
restore
*-------------- end example --------------------

-----------------------------------------
Maarten L. Buis
Institut fuer Soziologie
Universitaet Tuebingen
Wilhelmstrasse 36
72074 Tuebingen
Germany

http://home.fsw.vu.nl/m.buis/
-----------------------------------------


      

*
*   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/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index