Saturday, February 11, 2012

New version of phylosig

I just posted a new version of phylosig with a couple of minor updates.

First there was a bug (due to a misspelling) that caused the function to return NULL for convergence, even if it had in fact converged.

Second, a user pointed out that I had not specified what good starting values for the optimization might be. Let me say that I have tried to improve the default starting values for optimization in the phylosig(...,method="lambda",se=SE) model, and specifying start is optionally. However, if you do want to specify start, it can be provided as a vector length 2 containing the starting values for optimization for σ2 and λ, respectively. Thus, the first value needs to be on the valid range for σ2 - that is, greater than 0. The second value should be between 0 & the maximum theoretical value of λ (this is usually very close to 1). Note that this will only used in the models that use optim, which is basically the λ with sampling error model.

Direct link to the code for this new version is here. I'm also planning to build a new version of phytools soon.

No comments:

Post a Comment