I’m fitting a model using
brms. The chains settle on one of two peaks in posterior probability space Looking at the posterior samples using mcmc_pairs allows me to see that one of the peaks is substantially more probable than the other. As a result, I’d like to simply remove the chains that converge on the lower peak and was hoping to identify them visually by labeling the posterior samples based on each chain.
I’ve looked into the
pairs_style_np documentation, but don’t see an obvious way to color code the samples by their chain ID.
Does anyone have any suggestions on how to achieve this? It might be obvious if you’re competent in the grammer of graphics behind ggplot2, but I must admit I’m a conceptual novice.