Graham:
In the paper by Gelman & Pardoe (see link)
several R-square like measures are proposed for each
level in a multilevel model.
http://www.stat.columbia.edu/~gelman/research/published/rsquared16.pdf
Ernst Linder.
Roberts, J. Kyle wrote:
Graham,
Don't report any R^2 statistic for any mixed effects model. Snijders &
Bosker and Hox give lengthy discussions on why not to. I am presenting
a paper at AERA this year on possible R^2 measures for 2-level linear
models, but I assure you that it is not easy, nor straight-forward. If
you look on my website at:
http://www.hlm-online.com/datasets/education/
Dataset 1 actually gives an example where adding a covariate increases
the variance at level-2, which would mean negative variance explained
(something not possible in OLS). Typically, an R^2-type metric makes
little sense in HLM/mixed effects models.
I hope that this helps,
Kyle
***************************************
J. Kyle Roberts, Ph.D.
Baylor College of Medicine
Center for Educational Outreach
One Baylor Plaza, MS: BCM411
Houston, TX 77030-3411
713-798-6672 - 713-798-8201 Fax
jkrobert@bcm.edu <mailto:jkrobert@bcm.edu>
***************************************
------------------------------------------------------------------------
From: s-news-owner@lists.biostat.wustl.edu on behalf of GRAHAM LEASK
Sent: Sat 2/18/2006 5:06 AM
To: s-news@lists.biostat.wustl.edu
Subject: [S] Mixed effects models and R2
With mixed effects models in S plus is it feasible to describe models in
terms of how much of the variation explained? If so how do I gain an
estimate of how much of the variation is explained by the model (i.e. an
R2 figure).
I am using S plus 7.0 but have relatively little experience in using
this software.
--
****************************************************************
Ernst Linder elinder@unh.edu
Department of Mathematics and Statistics 603 - 862- 2687
University of New Hampshire Fax: 603 - 862 - 4096
Durham, NH 03824 www.math.unh.edu/~elinder
****************************************************************
|