Jump to content

Talk:Shrinkage estimator

Page contents not supported in other languages.
From Wikipedia, the free encyclopedia

Bessel's Correction is not a Good Example

[edit]

I think the example of Bessel's correction is incorrect here. Shrinkage is needed because the activity of exploratory regression analysis (optimizing the fit) tends to find estimators that explain more noise than is present in the true data process. Bessel's correction arises because there is an explicit formula removing one degree of freedom (the use of the sample mean rather than the population mean in the estimator of the variance) and not because of over fitting tendency in the data. For any reasonable size sample, Bessel's correction is in practice a small correction. — Preceding unsigned comment added by 69.142.244.49 (talk) 01:10, 18 April 2012 (UTC)[reply]

Bessel's correction is not mentioned explicitly. In this context the use of the divisor n+1 in place of the n−1 in Bessel's correction is an example of shrinkage, as was trying to be implied. So Bessel's correction was/is not being used as an example of shrinkage. Melcombe (talk) 09:06, 18 April 2012 (UTC)[reply]