Dear all,
sorry for the late reply. I was away for a couple of days and I am
surprised to see that my original post got so much attraction ;-)
Am 19.09.19 um 20:51 schrieb Sven Schreiber:
Am 19.09.2019 um 18:36 schrieb Stefano:
> dear all,
> sorry for arriving late - been busy marking (appalling) exams and
> projects over the last few days.
> Coming to the issue, I confess I fail to see the point of resampling
> without replacement.
> As far as the bootstrap is concerned, as far as I know resampling MUST
> be carried out with replacement.
I agree, but I think the "bootstrap" qualifier is important here.
There are other resampling techniques as well. Artur should probably
tell us what inspired his suggestion in the first place, but I would
speculate perhaps cross-validation with randomly drawn subsamples.
An important keyword seems to be "non-replacement sampling".
The new msample() function does all I need: Draw n items (rows) from a
matrix with r>n rows _without_ replacement -- ordering doesn't matter.
I am trying to replicate a simulation study written in R where R's
sample() is heavily used
(
https://www.rdocumentation.org/packages/base/versions/3.6.1/topics/sample).
Sven you're right, it would also be needed for cross-validation with
randomly drawn subsamples which should only be drawn once.
And that simple functionality I missed in gretl. So thank you, Allin,
for the new feature.
Also, R's sample() function does not support block-length
parametrization. This sounds quite a special task to me and makes things
really tricky as you explained, Allin.
Best,
Artur