[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

RE: st: RE: Outlier: Detection

From   "Nick Cox" <>
To   <>
Subject   RE: st: RE: Outlier: Detection
Date   Thu, 21 Feb 2008 12:37:52 -0000

Technique for cycling through categories is discussed at (e.g.)

I highlight (again) the very specific nature of this Grubbs procedure. 

I suspect that many will have been misled by general names like "outlier
test" and the existence of programs to think there is more white magic
here than there really is. 
The monograph by Vic Barnett and Toby Lewis on "Outliers in statistical
data" published by Wiley (various editions) makes it clear how many
procedures have been suggested, each 
it seems born with as least as many limitations as attractions. 

I like this definition from W.N. Venables and B.D. Ripley, Modern
Applied Statistics with S, Springer, New York, 2002, p.119. 

"Outliers are sample values that cause surprise in relation to the
majority of the sample." 

It is a very short step from that to noting that such surprise is a
function of the model contemplated and the subject-matter knowledge of
the researcher, and not an inbuilt characteristic of the data. 

Austin Nichols

If the issue is that there are more distinct values than -levelsof-
can handle, that is easily resolved, unless I am missing some finer

*   For searches and help try:

© Copyright 1996–2017 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index