Positive posterior density where prior is zero?

Estimation ML/Bayesian estimation
1

Hi all,

I have this weird-looking Bayesian estimation result of the simplest model. I put uniform prior on everything on purpose, it is supposed to give me zero density of posterior on points outside of my prior support yet turned out not to be the case for psi and sigma_eps.

Could anyone explain this?

uniform_prior
uniform_prior1135×331 10.8 KB

2

Usually, the chain has not yet converged. You should investigate the trace_plots. Regarding the zero: the black line derives from a kernel density smoother based on a Gaussian kernel. That may lead to small implied density at zero prior mass areas.

3

Thank you so much! That’s exactly what I was looking for.