0

I am currently reading the book called “Intro to Statistical Learning”. Here, it firstly generates new samples by simulation (1000) to get population mean and then estimates std.err. But then states that

In practice, we can not generate new samples from original population ...

and chooses “Bootstrapping”.

In this approach we obtain distinct samples by repeatedly sampling from the original (real) data set and this approach allows some observation to be repeated more than once in a given sample.

So, my question is that don't we follow the same procedure in the “bootstraping” because here we also somehow generate new samples from the data implementing some kind of simulation? What does it mean (the sentence quoted above)? Consequently, what is the difference?

Source -- The introduction to statistical learning, Chapter 5.2, page 188-189

MarianD
  • 1,493
  • 2
  • 8
  • 17
Davud Mursalov
  • 11
  • 2
  • 1
    Bootstrapping resamples your sample, so you will only get units that are already in your sample. You will never get any new values with bootstrapping. – user2974951 Jan 20 '21 at 12:38
  • Your quote is imprecise, it should be *"In practice, however, the procedure for estimating SE(ˆα) outlined above cannot be applied, because for real data we cannot generate new samples from the original population. However, the bootstrap approach allows us to use a computer to emulate the process of obtaining new sample sets, so that we can estimate the variability of ˆα without generating additional samples. Rather than repeatedly obtaining independent data sets from the population, we instead obtain distinct data sets by repeatedly sampling observations from the original data set."* – MarianD Jan 20 '21 at 12:49
  • Thanks for the useful explanation user297.. .I got it. – Davud Mursalov Jan 20 '21 at 13:45

0 Answers0