I have two general questions about competing risk regression with Stata18:
1) The presentation of stcrreg on the stata website (https://www.stata.com/features/overv...ks-regression/) says "Time-varying covariates and coefficients are allowed": does this means that one shouldn't care about varying coefficients and Schoenfeld residuals (I see that estat phtest doesn't work after stcrreg ), or just that it accepts vce() texp() subcommands, which it does?
2) I see that interpretation of the coefficients of time varying variables (particularly internal variables), is particularly troublesome (https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6916372/). One suggested solution is to use a "landmark model" (https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3702649/), which I am not sure to understand completely, but seems to consist in segmenting periods of time and testing each one using fixed covariates . Is there a way to do this with Stata? Or it is just a matter of building a different variable for each period, with a single value of the changing covariate, and stset the data in different ways?
thanks!
piersante sestini
1) The presentation of stcrreg on the stata website (https://www.stata.com/features/overv...ks-regression/) says "Time-varying covariates and coefficients are allowed": does this means that one shouldn't care about varying coefficients and Schoenfeld residuals (I see that estat phtest doesn't work after stcrreg ), or just that it accepts vce() texp() subcommands, which it does?
2) I see that interpretation of the coefficients of time varying variables (particularly internal variables), is particularly troublesome (https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6916372/). One suggested solution is to use a "landmark model" (https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3702649/), which I am not sure to understand completely, but seems to consist in segmenting periods of time and testing each one using fixed covariates . Is there a way to do this with Stata? Or it is just a matter of building a different variable for each period, with a single value of the changing covariate, and stset the data in different ways?
thanks!
piersante sestini
Comment