[R-sig-ME] Testing for significant intercept-slope correlation in nlme
khallg at unm.edu
Fri Aug 10 16:47:16 CEST 2012
Thanks Dave! The syntax you gave worked great for testing each aspect
of the random effect.
I also tested the models without the autocorrelation component and a
few of the fixed effects estimates had quite different magnitudes than
they did when autocorrelation was included (e.g., intercept of 2.0
without autocorrelation vs. 1.5 with autocorrelation). The direction
and significance of the effects didn't appear to change much though.
That makes me inclined to retain that component, with the hope that
it's doing more good than harm.
More information about the R-sig-mixed-models