[R-sig-ME] Calculating an Intraclass Correlation Coefficient or Repeatability estimate from lmer() output
amelie lescroel
amelie.lescroel at cebc.cnrs.fr
Wed Oct 14 14:24:44 CEST 2009
Dear all,
I am interested in calculating an Intraclass Correlation Coefficient (ICC) or Repeatability estimate from a mixed model output (lmer()). Lessels & Boag (1987) defined repeatability as the intraclass correlation coefficient based on variance components derived from a one-way anova:
r = among-groups variance components / (within-group variance components + among-group variance components)
What I would like to know is where/how to find these variance components in a lmer() output.
Specifically, we measured foraging efficiency (CPUE) of birds during 10 consecutive years. We have several measures per bird for each year and the same birds were measured over multiple years. What I would like to get is an estimate of the intra-individual consistency of foraging efficiency over time. And I thought that the ICC or repeatability would be this estimate (recently used in the same manner in Lang et al. 2009 Ecology 90(9): 2513-2523).
Let’s run a model with the ID of the birds as a random factor:
> (fm1 <- lmer(log.CPUE~(1|ID)))
Linear mixed model fit by REML
Formula: log.CPUE ~ (1 | ID)
AIC BIC logLik deviance REMLdev
-1144 -1126 575.2 -1158 -1150
Random effects:
Groups Name Variance Std.Dev.
ID (Intercept) 0.010292 0.10145
Residual 0.036602 0.19132
Number of obs: 3320, groups: ID, 341
Fixed effects:
Estimate Std. Error t value
(Intercept) 0.3484 0.0068 51.24
In this case, is the within-individual variance = 0.010 and the among-individuals variance = 0.037?
Thanks for your help,
Amelie
Amélie Lescroël
Seabird ecologist
URU 420, Université de Rennes I - Service du Patrimoine Naturel, Museum National d’Histoire Naturelle
263 Av. du Gal Leclerc
35042 Rennes Cedex
France
__________ Information from ESET Mail Security, version of virus signature database 4506 (20091014) __________
The message was checked by ESET Mail Security.
http://www.eset.com
More information about the R-sig-mixed-models
mailing list