I am trying to bootstrap standard errors for count data. Specifically, I want to get standard errors and confidence intervals on the total number of occurrences of an event. To fix ideas, here’s an example. Say I want to get an estimate of the standard error/ CIs for the number of tornados in Oklahoma per year, and I have a sample of tornados by county. For each bootstrap iteration b, I could: 1) cluster-sample counties with replacement and 2) calculate an estimate of the number of tornados for bootstrap sample b. I repeat 1) and 2) many times and use the estimates to get the standard error/confidence intervals.

I haven’t been able to find any resources that claim the process would be any different from a normal bootstrapping procedure, but I thought it might be worth asking around. Any ideas?

Looked for papers about count data specifically, only found papers that talked about bootstrapping for Poisson-based regression models.