I would like to seek your advice on a problem I encountered recently.
I bootstrap on a test that includes a Breslow estimate for survival data. Since these are meta-analyzes, I am bootstrapping on the trial (choosing randomly with replacement, $k$ trials among the $k$ trials, so the patients of one trial may appear few times).
This causes me to have about half of the data which are repeated. I noticed that the Breslow estimator fails to find an estimation for all the patients, and it stops at some point. Thus, I cannot calculate the bootstrapped statistic.
Do you think it's normal than the Breslow estimator fails at some point or another?
Do I have any way to prevent this maybe by changing the way my data are bootstrapped?