Hello there
I would like to use my survival analysis with weights. However stata won't allow me to run my univariate analysis logrank test.
Previously this was possible without the attweight variable
Stata returns:
option logrank not allowed
However, it is able to continue processing the below:
I don't particularly need the logrank test, as I have the graphs,
I found the below in stata help
If you have sample weights (if you specified pweights when you stset the data), you cannot run the log-rank or Wilcoxon tests. The Cox regression model, however, has been generalized to sample-weighted data, and Stata’s stcox can fit models with such data. In sample-weighted data, the likelihood-ratio statistic is no longer appropriate, but the Wald test based on the robust estimator of variance is
Questions:
1. Why can I not run log rank stat and why does stata help say 'it is no longer appropraite'
I would like to use my survival analysis with weights. However stata won't allow me to run my univariate analysis logrank test.
Previously this was possible without the attweight variable
Code:
stset survival [pw=attweight], failure(revised==1) //no problems here sts test gender, logrank
option logrank not allowed
However, it is able to continue processing the below:
Code:
sts graph, by(gender) stphplot, by(gender) nongegative nolntime stcoxkm, by(gender) //Subsequently with a plan to run stcox var 1 var2 var3, nohr
I found the below in stata help
If you have sample weights (if you specified pweights when you stset the data), you cannot run the log-rank or Wilcoxon tests. The Cox regression model, however, has been generalized to sample-weighted data, and Stata’s stcox can fit models with such data. In sample-weighted data, the likelihood-ratio statistic is no longer appropriate, but the Wald test based on the robust estimator of variance is
Questions:
1. Why can I not run log rank stat and why does stata help say 'it is no longer appropraite'
