Re: Corollary: N-P Silliness in Estimation Theory (was: Re: Unusual formulae for confidence intervals)


David Winsemius wrote:
"Reef Fish" <large_nassua_grouper@xxxxxxxxx> wrote in news:1164375456.941624.6580@xxxxxxxxxxxxxxxxxxxxxxxxxxxx:

The only person in this newsgroup who claimed to have used
the unbiased estimate for sigma was Jack Tomsky, who used it
in an indirect way, when he estimated the quantile of a
distribution.

Not necessarily the only one who ever used it, though.

That is a really frivolous nit-pick on an obvious hyperbole, though
as stated, the hyperbole is beyond any valid nitpick!

Jack was the ONLY person in this newsgroup "who claimed to have
used the unbiased estimate of sigma".

Can you name anyone else who made such a claim in this newsgroup?


http://biomet.oxfordjournals.org/cgi/reprint/69/3/635?ijkey=65cc8fb6d7aa6e5ebece1ccac5c0164e1a2aeeab

On p 639 of this 1982 Biometrika paper, Frank Harrell used the same estimator
for comparison with a distribution-free quantile estimator.

I am sure a few others did, after the paper in the American
Statistician
giving the unbiased estimator of sigma was around 1972. But for the
subject Harrell considered, the invariant MLE would have been a
much better choice as the estimator for comparison than an unbiased
estimator because it made Harrell's result WRONG everywhere
except in the unit he chose to investigate (because of the lack of
the invariance property of MLE), or as a measure-theorist would say,
his comparative result was valid on a set of measure ZERO. ;-)

But the POINT I think I have made, with adequate supporting reasons,
is the SILLINESS of over-stressing the importance (or any importance)
of the criterion of "unbiasedness", when the most commonly used
(and frequently used) estimate for sigma is the biased sample
standard deviation s.

Reef Fish Bob.

.

Relevant Pages

  • Re: Unbiased estimator of the parameter of a binomial distribution
    ... having a discrete uniform distribution on the interval. ... The idea is you have a known number of red and black marbles, ... and B black marbles, the maximum likelihood estimator of p, call it p* ... R / k an unbiased estimator. ...
    (sci.stat.math)
  • Re: Unbiased estimator of the parameter of a binomial distribution
    ... having a discrete uniform distribution on the interval. ... and B black marbles, the maximum likelihood estimator of p, call it p* ... R / k an unbiased estimator. ...
    (sci.stat.math)
  • Re: Q: Bootstrap
    ... I'm not a big user of bootstrapping myself so I hope I'm not introducing any mistakes here, nor am I certain to what extent my answer will relate to your particular issue, but here at least is one alternative explanation. ... A typical issue that bootstrapping addresses is the behaviour of an estimator when used on a given sample: ... We assume that we have n independent random variables X1,...,Xn drawn from a distribution P and an estimator Fof some parameter F_P describing the distribution P. ... Notice that I haven't mentioned the word "resampling" so far: I'm not doing resampling, I'm doing sampling from the empirical distribution. ...
    (sci.stat.math)
  • Re: Mean estimation problem
    ... robust M estimator can be used, ... Box, George C. Tiao ... coming from an unknown (but normal-like) distribution by using the ...
    (sci.stat.math)