Hi
All,
I am trying to model
a mixed model where the variance component of the random effect is modelled as a
continuous mixture. As hyperdistribution for the variance components I use a
gamma distribution. In my application, there are typically a lot of small
variances and a few higher ones (thus a high CV of the gamma distribution). When
I analyse a simulated dataset with a high CV, I get the following
error:
'update error for
node ... algorithm UpdaterSlice.Left error can not sample node too many
iterations'
This can be avoided
when truncating the gamma distribution for the very small values (e.g.
var~dgamma(alpha,beta)I(0.00001,)). However, this results in an underestimation
of the CV of this distribution, which is of primary interest to me. Is there
another way around this?
all suggestions
welcome
thanks
Stefan
Prof. Dr. Stefan Van Dongen
Department of Biology
- Group of Evolutionary Biology
University of Antwerp
Groenenborgerlaan
171
B-2020 Antwerp
Belgium
personal page:
http://www.ua.ac.be/stefan.vandongen
email: [log in to unmask]
Tel: + 32 (0)3
265 33 36
Fax: + 32 (0)3 265 34 74