Dan Warren
Dan Warren
Everything's working with viridis ramps now, but I'm thinking at some point we should sit down and figure out a consistent scheme for which color ramps go with which sorts...
https://arxiv.org/pdf/1903.08008 https://bookdown.org/content/4857/markov-chain-monte-carlo.html These seem like a very useful alternative visualization of MCMC outputs, and fairly easy to implement. It's basically just ranking all estimates for a parameter across all chains,...
The p values on these tests are all using the correct empirical values, but the plots that are spit out seem to be using the training AUC in geographic space...
https://nsojournals.onlinelibrary.wiley.com/doi/10.1111/ecog.07285
I've managed to chase this down to the fact that the invisible(capture.output()) bit is not returning a spatRaster, but need to figure out how to work around that.
https://www.r-bloggers.com/2025/09/key-improvements-in-shapviz-and-kernelshap/ They've got some built-in stuff to continue doing permutations until you reach convergence, which could be a good way to deal with the instability of permutation importance scores (getting...
Just like the current response plots but take the (weighted?) average prediction of a set of models across the range of each predictor.