Unit tree with arbitrary parameters

[richfitz]

Issue by richfitz from Tuesday Oct 01, 2013 at 06:05 GMT
Originally opened as traitecoevo/modeladequacy#21

---

It should be easy enough to write a method based on the fit.mle ones that takes a vector of parameters and makes a unit tree. This is easiest for BM with one parameter, but given we have a standard list of names, we could require (or assume) those.

[richfitz]

Comment by mwpennell from Thursday Oct 03, 2013 at 06:20 GMT

---

just saw this

assuming you mean for the simple models (multipeak OU and auteur models would be VERY hard)

like
as.unit.tree(tree, data=data, pars=c(sigsq=1, alpha=0.5))

that sort of thing? sure we could do this. do you think this would be useful??

[richfitz]

Comment by richfitz from Thursday Oct 03, 2013 at 06:27 GMT

---

I think that there are limited cases where this could be useful; there is one such case in the "BM tree rescaling worked (diversitree, mcmc)" test in test-unit-tree.R (see lines 64-67 here, which could be replaced by as.unit.tree(lik.bm, coef(samples.bm)[idx,] if we allowed something like this. Not sure if it's worth it, but not too much work either.

Yeah, for things like multipeak OU and auteur models it could be a bit harder, but with a likelihood function at least the order of the arguments is well defined.

[richfitz]

Comment by mwpennell from Thursday Oct 03, 2013 at 06:33 GMT

---

i see. if you feel like writing something like and think it would be useful, go ahead and give it a go.