I think I'm not getting the reasoning behind the bootstrap distribution, so I was wondering if anybody could clarify it for me...

This is an example from the textbook:

We have an original sample of size n=50. Then the textbook says that we can use a software in order to draw 1000 resamples of size 50 from the original sample. And then we can use the means of these resamples for the bootstrap distribution.

This is what I don't get: if our original sample is of size 50, and if we want to get a resample of size 50, then we're basically reusing every single thing of the original sample, right? Is that correct, or am I missing something? Because this isn't making a lot of sense to me...why would we take the exact same sample over and over again and then compute the mean, which will be exactly the same as the original sample's mean. So I'm probably missing something but I'm not sure what.

Thanks in advance

  • 2
    $\begingroup$ you sample with replacement - that is, a subject can appear more than once in the bootstrap sample. Example in R;x<-1:10; sample(x,10,replace=T) $\endgroup$ – user20650 Apr 17 '13 at 22:01

For your application, the key to the bootstrap is resampling with replacement. Each time, you draw some observations more than once and others not at all.

However, there is such thing as a subsampling bootstrap that involves taking smaller samples without replacement from the original sample.


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.