mode not at the highest point of density curve

[mindint]

Dear all,

I found that in many cases the posterior mode (green line) is not located at the highest point of the posterior density curve (see the attached file). Why is that and does it imply that the estimation results are unreliable?

Thanks

[jpfeifer]

Looking at this, I am pretty sure your mode-finding was not successful. This in itself is not a problem if the MCMC that you subsequently ran did converge to its stationary distribution. But this, due to using the incorrect mode, can take quite some time. Judging from the look of the posterior, this was not yet the case. You need more draws. Let me guess: you used the standard mode_compute=4? If yes, try e.g. 9.