Page 1 of 1

Mode_check plots

PostPosted: Mon Jun 04, 2012 3:15 pm
by Steven
Hello everybody,

I am doing a bayesian estimation with using the mode_check command to plot the posterior densities. I have some questions concerning the curves:

1. Is it important that the green curve (log-lik) is almost exactly the same as the blue curve (log-post). In some cases it is almost the same, in some cases there is a flat green one (which means that there is no information in the data), and in some cases, the green one is a little bit shifted. What does the last one mean? Is it a problem when the two curves are not overlapping?

2. In most plots I have seen that the blue curve has the shape of a smiling profile (U-shape). In my plots the curves are exactly the opposite (like an inverse U-shape) with a negative sign. The mode (sky blue vertical line) coincides always with the maximum of my (inverse U) plots. What does that mean or what is the problem within my estimation?

Thank you very much in advance!!

Best
Steven

Re: Mode_check plots

PostPosted: Fri Jun 08, 2012 11:42 am
by jpfeifer
1. The difference between the posterior and the likelihood tells you how important your prior is.
2. The plot should look like the attached one. The vertical line should be at the highest point of the concave posterior function (the blue one).

Re: Mode_check plots

PostPosted: Mon Jun 11, 2012 3:41 pm
by Steven
Thank you so much for your quick reply!

The plots looks like mine, so I guess everything is allright.

Best,
Steven