Skip to content
Prev 19921 / 20628 Next

Problem specifying uncorrelated random intercepts and slopes for a multi-df covariate

Thanks Jo?o. That's a cool page with all the illustrations, although I
still can't claim to quite get it.

I'd assume that in order for the random slopes to correlate with the
intercepts, they would first have to be estimated at some non-zero value.
THEN I see how a correlation could arise and how the slopes could perhaps
be pulled towards that correlation. In the case at hand though, we have
random slopes that are essentially zero. Hence, I don't see how a
correlation with the intercepts could possibly arise. And without such a
correlation, I don't see how the slope estimates could be pulled towards
one.

I wonder if I'm missing something.

Best,

J



pe 5. elok. 2022 klo 17.57 Jo?o Ver?ssimo (jl.verissimo at gmail.com)
kirjoitti: