nestedness of BiSSE within MuSSE/MuHiSSE

[GinaMCalabrese]

Hello, I am working on a project with character data that could be coded as binary (e.g. migratory, non-migratory) or as having three states (non-migratory, short distance migrant, long distance migrant). I fitted some BiSSE models to the binary data and some MuSSE and MuHiSSE models to the multi-state data. My understanding was that the BiSSE models are nested within the MuSSE and MuHiSSE models, and the MuSSE models would be nested within the MuHiSSE models. The likelihoods among the models do not make sense in light of this. Specifically, likelihoods should be lower for simpler models nested within more complicated models (i.e., mod1 is nested within mod2, and mod1$loglik < mod2$loglik). However all the MuSSE and MuHiSSE models (mod3-mod7) have lower likelihoods than the BiSSE models despite being more complicated (See mod.compare.table object in the attached script/workspace).

Am I incorrect about the nestedness of Bisse within MuSSE/MuHiSSE, did I set up these models incorrectly? Or is there something else I’m misunderstanding? Many thanks!

A note on files:
R_script_for_github.R is the code. It loads the following input files: character_data.RData (dataframes for the binary and mutli-state character data) and my_tree.tre (the phylogeny).
All R objects produced by the script (e.g. the hisse models) are saved into git_example_workspace.RData, so you can load this in order to save time by not having to fit all the models yourself.

files_for_github.zip

[thej022214]

Do you mind sending me an email (not with the questions -- I just refer to here) so that we can communicate offline? I think it might be easier that way.

Best,
Jeremy