competing risk without covariates

[Javier Roberti]

I want to compare the risk of dying of a specific cancer versus "other causes" in a large sample. Compare reason a versus b. I am trying to use stcrreg command but it requires covariables in the model, which I don't want to include. Is it possible to run this command without covariables? Is there a specific setting up that permits comparing two risks without covariables? When not entering covariables, it results in an empty model.
Thanks

[Paul Dickman]

If you don't want to model the effect of covariates then I would suggest using -stcompet- (available from SSC).

If you really want to model, then you could fit a flexible parametric model (-stpm2-) and estimate the cumulative incidence using -stpm2cif-. Both commands available from SSC. I can't guarantee it works with a null model but my guess is that it would; run one of the test examples first. See https://www.stata-journal.com/articl...article=st0298 for details.

[Javier Roberti]

Thanks Prof Dickman
I have tried following your instructions, even entering my data on the variables provided in your file and stata shows the same error.
No late entry detected - p is estimated using the actuarial method
using data not sorted
r(5)
Stata works ok with your data but not with mine. Stata v14.2

[Paul Dickman]

Is it possible you have put this in the wrong thread. That error message looks to be from -strs-. The command I suggested is -stcompet- (by Enzo Coviello).