Hi I am using metrop in MCMC library. Since some times the prior density is 0
and
the log prior density would be –Inf, I ask the return value for loglikelihood
equal to Min.log, which is defined as log(.Machine$double.xmin)-400, whenever a
–Inf occurs to the logdensity. However,
I noticed if two Min.log’s happen in a row, the sampled parameter value for the
second Min.log is treated as accepted.
As a result, the parameters result in 0 prior density are included in my
results, which I do not want to do. How
should I deal with it to avoid the acceptance of these iterations in MCMC?
Thanks for your help!
[[alternative HTML version deleted]]
