Re: Q: Bootstrap

Am 13.07.2007 18:35 schrieb Einar Andreas Rødland:

Notice that I haven't mentioned the word "resampling" so far: I'm not
doing resampling, I'm doing sampling from the empirical distribution.
Sampling from the empirical distribution just happens to be the same as
resampling with replacement from the original sample.

If you were to do your resampling without replacement, your resampled
X1*,...,Xn* would not be independent, so you are not really testing the
behaviour of F(X1*,...,Xn*) under conditions comparable to those assumed
for X1,...,Xn: i.e. independence. In fact, you'd have to make a smaller
resample, otherwise you'd end up invariably picking all X1,...,Xn.

Hi Einar -

now I'm beginning to get a clue... These arguments sound
very good and convincing, I'll think about them in more
depth now.

Thanks -

Gottfried

----

Gottfried Helms, Kassel
.

Relevant Pages

  • Re: ADC clock drift
    ... But the sampling clock is dead-on accurate at ... rate to virtually any amount, and that card locks ... The resampling is done in the DSP and obviously the chip did not miscalculate the sampling rate. ...
    (comp.dsp)
  • Re: Frequency/Sample rate
    ... of the information in the waveform. ... sampling at a fixed frequency any other ... than the fundamental of the wave. ... increased number of artifacts created when resampling. ...
    (rec.audio.tech)
  • Re: Frequency/Sample rate
    ... any waveform requires at least two samples per wave. ... sampling at a fixed frequency any other ... increased number of artifacts created when resampling. ...
    (rec.audio.tech)
  • down and up sampling
    ... for resampling pb ... all what i want is to combine every 3 pixel of spot to form only ... one(it is called down and up sampling). ...
    (sci.image.processing)