[Statlist] Talk on Statistics

Cecilia Rey rey @end|ng |rom @t@t@m@th@ethz@ch
Wed Apr 1 10:44:42 CEST 2009


  We are glad to announce the following talk
***********************************************************
*
Tuesday, April 7, 2009, 15.15-17.00 HG G 26.3*

by Carolin Strobl, Universität München, Institut für Statistik/
/

//
Title:
Concepts of Variable Importance

Abstract:
Random forest variable importance measures have become a popular
means of variable selection in many scientific fields (cf., e.g., Lunetta 
et al., 2004) because they can be applied to “small n large p” problems 
potentially involving complex interactions. The properties of the 
original random forest variable importance measures (Breiman, 2001) are 
compared to a new, conditional variable importance measure for random 
forests (Strobl et al., 2008) as well as more traditional approaches for 
the assessment of variable importance.

It turns out that – especially in the presence of correlated predictor 
variables – different characteristics of variable importance measures 
make them more or less well suited for the objective of variable 
selection, whereas the question “Which one
is the right measure?” remains a philosophical one.


The abstract is to be found under the following link: 
http://stat.ethz.ch/talks/research_seminar

-- 
ETH Zürich
Seminar für Statistik
Cecilia Rey-Lutz, HG G10.3
Rämistrasse 101                    
CH-8092 Zurich		                      	
mail: rey using stat.math.ethz.ch    	  		
phone: +41 44 632 3438/fax: +41 44 632 1228




More information about the Statlist mailing list