Commit 1d7a3a25 authored by Frédéric Santos's avatar Frédéric Santos

Update org file

parent 4f086df3
......@@ -558,4 +558,4 @@ For all those reasons, outlier detection is strongly user-dependent, and the str
The focus of the present article was on outlier detection, and not outlier management in a broad sense. The problem of kwowing what to do with the individuals that are detected as outliers is extensively covered in cite:leys2019_HowClassifyDetect. However, numerous robust methods have built-in way to handle outliers, and do not need a controversial manual exclusion. This article focused on robust correlation and regression methods, but most popular methods do have a robust equivalent which offers a valuable alternative for "contaminated data". Among other examples, robust principal component analysis citep:candes2011_RobustPrincipalComponent or robust estimation and hypothesis testing citep:wilcox2012_IntroductionRobustEstimation can be cited. Within the field of robust estimation, winsorization---i.e., replacing all the values exceeding a given threshold $t$ by the value $t$ itself---or trimming---i.e., removing a given percentage of the most extreme values in both directions---could be valuable tools in archaeology, and would offer some new ways to deal with outlying values.
* References :ignore:
Markdown is supported
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment