I'm trying to run a model using mixture of
normal distributions in WinBUGS and I would like to know if somebody how
to
consider the restriction in the distribution's means in order to avoid
the
permutability problem.
I have seen an example in Volume II in the Help of WinBUGS (Eyes: normal
mixture model"), in which a positive amount is added in the means
expression in order to differentiates the means of the distributions
that
compose the mixture. But in this example, there is no covariate. Is it
possible
to maintain this structure in a models with covariates?