[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]
st: R: STCOMPET
st: R: STCOMPET
Tue, 8 May 2007 15:12:14 +0200
I am using the program stcompet from Enzo Coviello, which I find
extremely useful. However, I am interested to calculate the confidence
intervals for the event of interest and for each competing event
separately. Does anyone know how to do that?
Thanks Peter for using -stcompet-.
I believe that your question is addressed in the SJ paper where -stcompet- is presented.
The example is:
use prostatecancer, clear
stset time, f(status==1)
stcompet CI = ci hilim = hi lowlim = lo, compet1(2) compet2(3) by(treatment)
Three new variables have been created: CI, hilim, and lowlim for the cumulative incidence and its high and low confidence bound. The estimates relating to a specific event are recorded at each time when this failure type occurs ..., two new variables are then generated to distinguish the estimates between treatment arms:
gen CI_tr = CI if treat
label var CI_tr "Treated"
gen CI_untr = CI if !treat
label var CI_untr "Untreated"
gen hi_tr = hilim if treat
gen hi_untr = hilim if !treat
gen lo_tr = lowlim if treat
gen lo_untr = lowlim if !treat
And for each event:
gen hi_tr_ev1 = hilim if treat & status==1 // hi_limits for treated and event of interest
gen hi_untr = hilim if !treat & status==1 // untreated
gen hi_tr_ev1 = hilim if treat & status==2 // hi_limits for treated and competing event type 2
gen hi_untr = hilim if !treat & status==2 // untreated
and so on.
If you have some further questions I can send privately the do file working the example described in the paper.