search

rmcminds / stan_models

collection of models written in Stan, with data wrangling code to run the models on real data
1 stars 2 forks source link

allow OU parameter to vary for different interactions #25

Open rmcminds opened 5 years ago

rmcminds commented 5 years ago

to really ask whether one clade of microbes is 'more cophylogenetic' than the background, that clade's branch lengths shouldn't just be scaled linearly, but should give more influence to the older nodes, such as by decreasing the strength of the OU attraction. so if the microbe tree has an overall OU attraction strength of 10, perhaps the strength is only 4 within the endozoicomonas group, meaning the ancestor of endozoicomonas entered a new mode so that its descendants are no longer pulled toward the overall mean as strongly.

rmcminds commented 5 years ago

leads to interesting interpretation of mechanics: that is, the realtime rate of evolution doesn't actually change, it's just that historical evolution sticks around longer. so microbes may always be host specific, but since they horizontally transfer so often, they erase their histories, but a lineage with reduced OU is no more host-specific than others, it just doesn't undergo the same amount of host swapping??

rmcminds commented 5 years ago

would need very strong data to be able to distinguish between a clade generally increasing rate of evolution vs clade reducing OU, I think. how do my ultrametric node transformations play into this? if a node is simultaneously 'moved' toward the tips and also has increased overall variance, it would approximate the reduction of OU attraction...

rmcminds commented 5 years ago

or perhaps better yet, keep the 'downstream' OU effects the same, but allow the accumulated effects to take the role of a new center. i.e. in an implementation that's calculating ancestral states, stop diminishing the effects of upstream effects and lock them into place.