RE: st: RE: generating data from AFT survival models with frailty. Feiveson, Alan H. (JSC-SK311) (Wed Mar 16 11:10:17 2011)

[Nick Cox <n.j.cox@durham.ac.uk>]

Pedantic comment here:

The code  

	tempname vlambda
	tempvar `vlambda'

creates a tempname which would be later referred to as `vlambda' and _another_ tempname (identified as a variable name) which would be later referred to ``vlambda'' (except that is not done). 

Antoine can do exactly what he intends (and nothing else) with the single line 

	tempvar vlambda

Nick 
n.j.cox@durham.ac.uk 

Antoine Terracol

you can generate a gamma variate with -rgamma()-

to draw from a Weibull distribution, you can use this bit of code:

cap prog drop draweib
program define draweib
syntax newvarlist  [if] [in] , LAmbda(string) ALpha(string) [double]
tokenize "`varlist'"
while "`1'"!="" {
     tempname vlambda
     tempvar `vlambda'
     gen `vlambda' = exp(ln(`lambda')/`alpha')
     g `double' `1' =((log(1/uniform()))^(1/`alpha'))/`vlambda' `if' `in'
     mac shift
     }
end



so, to simulate data from a weibull with gamma frailty, you could

clear
set obs 1000
generate frailty=rgamma(1,1)
drawnorm x1 x2
draweib t, alpha(.75) lambda(exp(1+x1+x2)*frailty)

and estimate the model with

stset t
streg x1 x2, dist(weib) frailty(gamma) nohr


You could of course censor the durations as you see fit, and change the 
-stset- declaration accordingly.

To draw from an exponential, jut use alpha(1) in -draweib-

Since the weibull/exponential are both PH and AFT you can choose 
whatever metric you like, , the coefficients in the AFT metric beeing 
equal to -(coeff in PH metric)/alpha


I'm sure it's also possible to simulate from log-normals and 
log-logistics with frailty as well, but I have no code to implement it...

Antoine


On 16/03/2011 19:58, Albert Navarro wrote:
> Thank you very much, Alan. Do you know any paper/document where it is implemented in a particular model?

*
*   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/