survival analysis accounting for survey weights

[Xiaochen Zhang]

Hi all,

I'm interested in all-cause mortality and cancer-specific mortality using survey data. I'd like to plot the survival function by group, and use stcrreg to address the competing risk. However, with survey weights, these commands could not be used. Does anyone have any idea handling survival analysis with survey weights?

Thank you for any inputs!
Xiaochen

[Steve Samuels]

Add:

 [pweight=surveywt]

and

vce(cluster psu)

to the stset statement. The latter can instead go into the stcrreg statement; I don't know if it makes a difference. If the survey strata are groupings like like "region" or "urban/rural", add them as predictors to the stcrreg model.

[Xiaochen Zhang]

When I add the vce(cluster psu), Stata gives me "invalid vcetype"

[Steve Samuels]

Sorry, you're right: vce(cluster psu) is not an option for stset. However it does work with stcrreg.