Skip to content
Prev 5269 / 20628 Next

Logistic regression with spatial autocorrelation structure

On Mon, Feb 7, 2011 at 10:58 AM, <Dale.W.Steele at gmail.com> wrote:
Didn't I answer 2 in my previous message in the part about
(Intercept)   priorprob  genderMale         wbc
 -8.2886183   0.0502540   1.1870539   0.3412998
[,1]
[1,] -2.299365
[,1]
[1,] 0.09117557

It is certainly possible to automate the process a bit more with a
predict method but, for the time being, I think that should be the way
to go about it.

If you have a simple random effect for doc, as in model m2, then to
get a prediction for a specific physician you add the random effect
for, say, doc == "18", to the value of eta before transforming it.
You can get the conditional modes of the random effects with the ranef
extractor.