search

cjcarlson / embarcadero

🌲🌉 Species distribution models with Bayesian additive regression trees
49 stars 11 forks source link

Spatial covariates #18

Closed grwhumphries closed 4 years ago

grwhumphries commented 4 years ago

Question more than an issue - but wanted to know how the algorithm handles spatial covariates - i.e. with INLA you would used SPDE to create a spatial smoother. Does BART have a way of incorporating those kind of spatial covariates that deal with spatial autocorrelation?

cjcarlson commented 4 years ago

Sorry for the delay! There's not an inherent solution built in, but you could use x and y as variables. There are more advanced ways to constrain possible variable interactions in dbarts, but I don't really know them yet, and haven't built them in here. But the short answer is that like most other ML methods for SDMs (e.g., BRT or MaxEnt), it's not explicitly spatial