I am trying to partition my data so that I can assign different models to different sites. In effect, I change from

LikelihoodFunction likelihood = (allsitesfilter, allsitestree);

To instead use:

LikelihoodFunction likelihood = (filter1, tree1, filter2, tree2, ...)

Eventually, each tree (tree1, tree2, ...) would get a different model. However, the trees would all have their branch lengths constrained to be equal, so there are no additional parameters in the second case if all trees use the same model. For a sanity check, I first just tried doing this where each tree gets the same model. I believe in this case the same inference is being done in both cases.

However, the first case converges to a solution fairly quickly. But the second case is still running! Is there some reason the performance is so slow -- and if so, how do I speed it up?

Hi Sergei,

The output shows that the one with partitions does include one additional parameter -- the root branch length, Node0.t. I am not sure why adding partitions make the root node start being treated as another parameter. But in any case, this only increases the number of parameters from 40 to 41, and seems unlikely to cause the huge slowdown.

You can see in the output from the slow (partition) case that there are lots of other parameters, but they are all constrained, so I don't think they should slow it down?

For some reason the message board is not allowing me to upload any attachments. I get the message:
"With this file to_post.zip (53 KB) you exceeds the maximum allowed upload-directory volume."
Even though my attachment is only 53 KB and the site says the limit is 1000 KB.

I can e-mail you the files since it won't allow me to upload them -- post here if I should do that.

Thanks,
Jesse