What is the minimum number of bootstraps to be done using bca non-parametric method

Question

I would like to use the boot() and boot.ci() functions from library("boot") for very large data sets(~50 000) with type="bca".

If R(number of bootstraps) is too small, it will give the following error:

Error in bca.ci(boot.out, conf, index[1L], L = L, t = t.o, t0 = t0.o,  : 
  estimated adjustment 'a' is NA

I wouldn't want it to be too large either.

What is a good number for R? I know it would depend on the size of the data.

Answer 1

You need a greater number of bootstraps, in general, for confidence intervals than for estimates. I have heard suggestions of 1000 bootstraps for estimates and 10000 for confidence intervals since computing power has increased in the last decade.

The boot() function nowadays has argument parallel which can run the bootstraps on multiple cores, which if compute time is a concern for you, you might look to use to reduce significantly the time it takes to run the analysis on data sets of your size.