bsvalidation "no observations r(2000)"

[Virginia Zarama]

Hello

I a running an internal validation for two clinical scores to predict intensive care unit (ICU) admission.
I am using the bsvalidation command. I had used it before for multivariable models and it works great, but for one of these two univariable models, it is giving me an error "no observations r(2000)".
The database has 6900+ patients, with no missing data on these variables.

Here is the code for the score that works fine (ROX). It is a continous variable from 0.65 to 47, nonnormally distributed (900+ unique values)

logistic uci ROX
bsvalidation, rseed(123) reps(50) adjust(bootstrap) eform graph

And here is the code for the score that is giving me the error (NEWS2): continuos variable from 0-19 (20 unique values), also nonnormally distributed.

logistic uci NEWS2
bsvalidation, rseed(123) reps(50) adjust(bootstrap) eform graph

If anyone knows how to correct this, please let me know.
Virginia

[Virginia Zarama]

Attached is a sample of my database with 200 observations and the dofile.

[Borja Fernandez-Felix]

Hello Virginia, the error is probably due to the fact that one of the risk groups has no events. You can try including the group() option and creating fewer groups, e.g., group(5).

bsvalidation, rseed(123) reps(50) adjust(bootstrap) eform graph group(5)

I have tested it with the sample you provided, and it works.