hi! i loved your video , even if my mother tongue is not english i was able to understand a 97% ! i would like to learn inference, and multivariate analysis in R , if u have any videos it could help me a lot!! greetings from italy by a Peruvian girl .lol.
In R, sample(x, replace=TRUE) returns in a vector of size length(x) (assuming x is a vector) samples of x drawn with replacement. The (default/standard) bootstrap method uses indeed sub-samples with the same size as the original sample.
I'm glad to help. In practice, if you have some computational constraint, you may use sub-samples of smaller size and adjust the "theory" (variance, confidence intervals etc) accordingly. I see no harm in that. The R code is like sample(x, size=n, replace=TRUE). About your last phrase, I think you mean that the bootstrap (bs) samples contain only values already observed in the original sample, which is true, but the values are re-used as they appear in the original sample (following the observed frequencies of smaller or greater values). Bootstrapping is equivalent to sampling from the distribution which puts 1/n probability mass on each x_i (this is the empirical distribution). It converges strongly with n to the true (unknown) distribution from which the original sample came from. This means that, as n grows, bs is like we were sampling from the true distribution function.
Hi, Jeff! Would you like to send history.R of your lesson? I need to test it. Thank you!
hi! i loved your video , even if my mother tongue is not english i was able to understand a 97% ! i would like to learn inference, and multivariate analysis in R , if u have any videos it could help me a lot!! greetings from italy by a Peruvian girl .lol.
Can you share with us the link to download the PDF ?
Hi Mr. Leek: Do these sub-samples which are drawn from the sample have smaller size or the same size as the sample?
In R, sample(x, replace=TRUE) returns in a vector of size length(x) (assuming x is a vector) samples of x drawn with replacement. The (default/standard) bootstrap method uses indeed sub-samples with the same size as the original sample.
Thank you very much. So x need not have the same size as the original sample. It seems the standard option implies making copies of the original data.
I'm glad to help. In practice, if you have some computational constraint, you may use sub-samples of smaller size and adjust the "theory" (variance, confidence intervals etc) accordingly. I see no harm in that. The R code is like sample(x, size=n, replace=TRUE). About your last phrase, I think you mean that the bootstrap (bs) samples contain only values already observed in the original sample, which is true, but the values are re-used as they appear in the original sample (following the observed frequencies of smaller or greater values). Bootstrapping is equivalent to sampling from the distribution which puts 1/n probability mass on each x_i (this is the empirical distribution). It converges strongly with n to the true (unknown) distribution from which the original sample came from. This means that, as n grows, bs is like we were sampling from the true distribution function.
Is this video from some online course? Great video!
I know it is pretty randomly asking but do anybody know of a good site to watch newly released tv shows online ?
@Juan Billy i use Flixzone. You can find it on google :)
@Henry Adrien Yup, I've been watching on flixzone for since march myself :D
@Henry Adrien thank you, signed up and it seems like a nice service :) I really appreciate it!!
@Juan Billy happy to help :D
Great Video! Thanks :)
Thank you so much
Thanks a lot