I am estimating variance components for a model with heterogeneous
error variances using Gibbs sampling. This is straightforward for
a model where we simply classify records as to which error variance
they represent, sampling from inverse chi-square distributions.
Assuming error variances change with time, it would be
preferable though to fit a variance function, e.g. a polynomial function of
time, to model changes in error variance with time.
Question is, how to do this ? what is the distribution of these polynomial
coefficients to sample from (assuming data are multivariate normal).
I would appreciate any pointers to publications dealing with this particular
problem - to be honest, I am looking for a 'recipe' at this stage rather than
in-depth theoretical treatment.
cheers,
karin.
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++
Karin Meyer, Animal Genetics and Breeding Unit,
University of New England, Armidale, NSW 2351, Australia,
Phone (+61) (02) 6773-3331 Fax (+61) (02) 6773-3266
=================================================================
Instructions for joining and leaving this list and remarks about
the problem of INAPPROPRIATE MESSAGES are available at
http://jse.stat.ncsu.edu/
=================================================================
