PROC GENMOD; class type county MODEL N_SURV/N_DIAG=ln_DIAG type
REPEATED SUBJECT=county
would be two lines of code in Stata:
generate double survival_proportion = N_SURV / N_DIAG
xtgee survival_proportion ln_DIAG, i(country)
If you had wanted to fit a logistic model similar to what Martin Weiss alluded
to, then you could have put the denominator variable in the family option of
xtgee:
xtgee N_SURV ln_DIAG, i(country) family(binomial N_DIAG)
Now, I have two questions:
How common is it in your field to put a variable (N_DIAG) on the left-hand side
of a linear model and its logarithmic transform (ln_DIAG) on the right-hand
side? I assume that it's done in order to make interpretation somehow easier.
If you have individual survival times, why not fit models specifically for
survival, instead of trying to fit linear models to proportions? I don't know
what SAS has specifically for frailty model in survival analysis (survival
models that involve, for example, patients within countries). Type -help st- to
see what Stata has to offer.
Joseph Coveney
*
* 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/
