Suppose I use the nonparametric bootstrap to fit a model as so:

  BucketFit <- fitdist(x,"weibull",  method = "mle", lower = c(0,0))
  BucketBoot = bootdist(BucketFit, bootmethod = "nonparam")

However, if I look at BucketBoot, there's a $fitpart component as well as CI component that gives me the distribution and CI. However, since I'm using nonparametric bootstrapping, these aren't actually used?

However, if I got to BucketBoot$estim I get bootstrapped values of parameters. So it still does a fit?



?bootdist tells you:


fitpart The object of class "fitdist" on which the bootstrap procedure was applied.

Thus the fitpart is just your previous fit: BucketBoot$fitpart == BucketFit non-parametric bootstrap here means that bootstrap samples are generated by resampling from the original data with replacement (in contrast to generation of bootstrap samples from the fitted model that happens with the default bootmethod = "param").

So to directly answer your questions: No, the fitted parameters are not used for the bootstrap in this case (Or only as a good starting point to speed up the bootstrap fits). And yes, it will do one fit for each bootstrap sample.

On a side note: the bootstrap CIs given by fitdistrplus are percentile CIs, you might want to compute basic bootstrap CIs instead.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.