Dear all
I am trying to fit a mixed model with a factor and a random effect on a slope:
y~F*x+...,random=~x
where F is a factor with 2 levels and x the covariate.
the random effects for the 2 levels of F should be equal so I am fitting the
model like:
ex.lme=lme(y~x+F+z*x+F*x-1,
random=list(group=pdBlocked(list(pdIdent(~F-1),pdIdent(~F*x-1)))))
In that case the random effect part of the summary is like:
Block 2: F1:x, F2:x
Formula: ~F:x - 1 | group
Structure: Multiple of an Identity
F1:x F2:x Residual
StdDev: 4.015003e-06 4.015003e-06 0.7636376
but the fixed part is like:
Fixed effects: y ~ z + F + z * x + F * x - 1
Value Std.Error DF t-value p-value
z -0.1012095 0.1174244 18 -0.861913 0.4001
F1 2.4558678 0.7743866 18 3.171372 0.0053
F2 1.4761337 0.7913741 18 1.865279 0.0785
x 0.0000006 0.0000040 705 0.158342 0.8742
z:x -0.0000010 0.0000007 705 -1.297161 0.1950
F2:x 0.0000030 0.0000027 705 1.082969 0.2792
I suppose that x value corresponds to the F1:x.
However, when I extract the coef() I get columns for x, F1:x, F2:x.
x coef is constant as if it was a fixed effect. I guess that F1:x is the random
and x the fixed part of the F1:x interaction so I should simply add these
columns for the final group-level coefficient.
Is that right?
Can I formally check or fix that to get a straightforward result?
Thank you!
Irene