Interpretation of correlations in multilevel anallysis

Question

In the following multilevel analysis in R (taken from here, page 57):

> Model.1<-lme(WBEING~HRS+G.HRS,random=~1|GRP,data=bh1996, control=list(opt="optim"))
> summary(Model.1)
Linear mixed-effects model fit by REML
Data: bh1996
     AIC      BIC  logLik
19222.28 19256.81 -9606.14

Random effects:
Formula: ~1 | GRP
           (Intercept) Residual
StdDev:
            0.1163900 0.8832353
Fixed effects: WBEING ~ HRS + G.HRS
            Value     Std.Error  DF    t-value  p-value
(Intercept) 4.740829  0.21368746 7282 22.185808 <.0001
HRS        -0.046461  0.00488798 7282 -9.505056 <.0001
G.HRS      -0.126926  0.01940357   97 -6.541368 <.0001

Correlation:
       (Intr) HRS
HRS    0.000
G.HRS -0.965 -0.252

...

How can the latter table (Correlation) be interpreted? What do these correlations mean?

score 1 · Answer 1 · answered Jul 06 '14 at 18:35

It's the correlation between the estimates of the true fixed effects, arranged in an ink saving way.

The values of the variable G.HRS are far away from zero. This explains the high correlation to the estimate of the intercept. If you don't like this, you can center G.HRS to mean zero.

Interpretation of correlations in multilevel anallysis

1 Answers1