[Rd] mixed effects model and r-squared

[Indermaur Lukas]

Heya

I am fitting linear mixed effects model in R and want to assess the model fit (with Animal number as random factor; repeated measures for Animals):

ts.model <- lme(LOG_FOC_MW ~ R_DN_SUM + ANIMAL + SEX+ YY, data = t.data, random = ~ 1 | ANIMAL, correlation=corCAR1(0.2, form = ~1 | ANIMAL ), method='ML', na.action=na.omit)).

Is there a possability to easly compute an R-square from the output of the model summary?

I would appreciate any hint.
Best regards

Lukas




???
Lukas Indermaur, PhD student
eawag / Swiss Federal Institute of Aquatic Science and Technology
ECO - Department of Aquatic Ecology
?berlandstrasse 133
CH-8600 D?bendorf
Switzerland

Phone: [phone]
Fax : [phone]
Email: lukas.indermaur at eawag.ch
www.lukasindermaur.ch

[Ben Bolker]

Heya
I am fitting linear mixed effects model in R [snip]
ts.model <- lme(LOG_FOC_MW ~ R_DN_SUM + ANIMAL + SEX+ YY, data = t.data,

random = ~ 1 | ANIMAL,

correlation=corCAR1(0.2, form = ~1 | ANIMAL ),

method='ML', na.action=na.omit)).

Is there a possability to easly compute an R-square

from the output of the
model summary?

I would appreciate any hint.
Best regards
Lukas

You probably need to repost this on r-help instead of r-devel;
it's not a "development" question.
The other bad news for you is that I suspect it may be
difficult to define r-squared uniquely for a mixed model.
The resid() command will give you
residuals, and you could take (1-resid()^2)/var(x) --
but how do you decide which var(x) to put in
the denominator ... ?

Ben Bolker