On Fri, 22 Nov 2002 22:20:26 +0900
Budana_Prijadi@fmi.com wrote:
>
> Special thanks to Dr. Stephane Heritier (Univ Sydney) who recommended "robust
> methods" and more.
> Also to Eric Zivot and Bert Gunter (Merck) who suggested the same methods.
>
> I'd better gear up for "modern methods" then. But I feel it is difficult to
> grasp the concepts of robust methods without knowing the concepts of outlier
> identification and extreme data points. A classroom issue?
>
> Budana
I think robust methods (including nonparametrics) are the way to go. I find
that labeling something as an "outlier" is arbitrary and leads to unscientific
behavior on the part of some researchers (e.g., deleting points that just
lowered the R^2).
Frank
--
Frank E Harrell Jr Prof. of Biostatistics & Statistics
Div. of Biostatistics & Epidem. Dept. of Health Evaluation Sciences
U. Virginia School of Medicine http://hesweb1.med.virginia.edu/biostat
|