Jump to content

Talk:Redescending M-estimator

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

Scale?

[edit]

It seems like redesigning M-estimators assume a certain window size. If I know that I expected Gaussian data with σ=1, I could see using a redescending M-estimator that excluded points outside of five or six standard deviations, since that would start to be more likely to be an outlier than to be accurate data, but if I don't know σ a-priori, how would I decide what window size to use? —Ben FrantzDale (talk) 15:09, 15 October 2008 (UTC)[reply]

yes this is the most important potential drawback of redescending M-estimators - I have expanded on the disadvantages Nick Mulgan (talk) 01:41, 26 September 2021 (UTC)[reply]

Notation

[edit]

This article mainly uses a capital Ψ, but one ψ, while the [M-estimator] article uses the lower case ψ: is there a standard notation for this? Melcombe (talk) 15:25, 15 October 2008 (UTC)[reply]