Skip to content
Prev 130 / 20628 Next

fixed effects correlated with the intercept

Is'nt this what might be expected.  Center the covariate about
its mean and, depending on the detailed variance-covariance
structure, the correlation may well reduce to zero.

Check this out with a model created using lm(), where it is
easier to follow the detail.  If you write the model  y = a + b(x -  
mean(x)),
the estimates of a and b are uncorrelated.  If x is not centered,
then you have
y = a - b mean(x))] + bx = adash + bx.

Then
adash = a - b mean(x)
involves b, and is clearly correlated with b. By making mean(x)
large enough or small enough, the correlation can be made
arbitrarily close to -1 or 1, respectively.

What do you mean when you say "I have two covariates that I
consider to be controls in my model." Do you mean that these
code for observations that you are treating as controls?  Or
what?

John Maindonald             email: john.maindonald at anu.edu.au
phone : +61 2 (6125)3473    fax  : +61 2(6125)5549
Centre for Mathematics & Its Applications, Room 1194,
John Dedman Mathematical Sciences Building (Building 27)
Australian National University, Canberra ACT 0200.
On 24 Mar 2007, at 6:07 AM, Austin Frank wrote: