Including a multivariate constrained improper prior in a hierachical model

Unfortunately for me I’ve already read the doc you linked and it hasn’t been of much help. CAR is fairly easy to implement since the prior for eta has a proper Normal distribution. The ICAR model im dealing with is much more of a challenge since the prior is improper, and implementing that sum-to-zero constraint is my biggest concern right now.

Thanks for your time, you can always come back and offer input when you can.