Closed inti closed 11 years ago
You use the Normalized Hamming distance of the sates between successive sampling iterations in your article. Do you think that is in general a good estimate? Inti
Take a look at this writeup to see some good alternative mixing tests.
thank!
Hi Matt, After your great help on issue #5 it is fast enough for me to test it on my application. i wonder if you have any suggestion on how to test for convergence. Would you test for convergence of the parameters (all or some important ones)? Could one use the convergence of the state assignment to different data points?
Any suggestion is welcome. I know one would normally go in line with the first of above but if there is a alternative to check convergence of 10s of parameters I would be happy.
BW, Inti