Posterior distribution

[jolee]

Hi everyone,

I am getting very narrow posterior distributions (90% interval <0.0001) and the trace plots seem to move only around the starting points.

Why am I getting this problem and how can I fix it?

Thanks,
jolee

[jpfeifer]

Your MCMC did not correctly sample from the posterior, potentially because the MCMC jumping covariance or mh_jscale were too small. Did the mode-finding work? How was the acceptance rate?

[jolee]

Yes, jscale is small (0.011666) and acceptance rates are 32.66% and 32.74%.

Mode finding is not working at all and most of the likelihood functions are linear. Is that the matter of my data?

Thank you so much for your help.

[jpfeifer]

It is hard to say. Please recheck your model and the observation equations and try different mode finders. You also might want to try the

Code: Select all

MCMC_jumping_covariance

option.