search

eggplantbren / TimeBombs

GNU General Public License v3.0
0 stars 2 forks source link

Are the DNest chains converged? #9

Open drphilmarshall opened 9 years ago

drphilmarshall commented 9 years ago

In 10% of simulations, the true time delay is in a region of very low posterior density. This does not smell right - perhaps the chains are not converged? @narumbaugh, in one or two of these problem cases, does changing the DNest control parameters improve the sampling and provide non-zero posterior density at the true point? Can we auto-test the chains for convergence, @eggplantbren ?

eggplantbren commented 9 years ago

This could well be the case (stuck in a local mode, for example -- this should happen less with DNest than with many other methods, but it still happens).

drphilmarshall commented 9 years ago

@rumbaugh's going to make some cornerplots, showing all hyper-parameters given each dataset. That may give you some clues as to whether DNest is getting stuck, perhaps?

On Mon, Jul 6, 2015 at 6:10 PM, Brendon J. Brewer notifications@github.com wrote:

This could well be the case (stuck in a local mode, for example -- this should happen less with DNest than with many other methods, but it still happens).

— Reply to this email directly or view it on GitHub https://github.com/eggplantbren/TimeBombs/issues/9#issuecomment-119039522 .