James Boyko

@jboyko.bsky.social

609 Followers 619 Following

a potential nasty consequence. these are profile likelihoods for a default model (b/d) and a more model (a/c). (b/d) is what you'd get fitting a standard Pagel model with ARD. You get these horrible likelihood ridges for parameter estimates (e.g., b. theta4). But for the same dataset, there is a slightly non-standard model (non-standard in the sense of it's not ER/SYM/ARD) that has really nice statistical behavior