Not an issue, just a comment - with a kudos note!

[TiagoAMarques]

I really trully appreciated your post, very insightful, and the fact that it has all the code to obtain the relevant data and run it makes it so useful. will use it in class. I really liked the fact that you used the bivariate smooth to soak up some of the spatial correlation, while this is often done, I have not seen it much (I work in the ecological statistics realm) to be explicitly stated as being a way to do that, account for the spatial correlation that otherwise would remain. Kudos for making this freely available to others. A final comment, when looking at your final plot, which you deem as the most approopriate for the data at hand, it looked at first quite weird, because of the ruggedness of the lines. I guess this makes sense since, apart from the pannels by state, the plot is not on geographical space (since the x axis is percent depression) while the model space is explicitly geographic. I was wondering therefore, what might be a bether way to visualize that result that does not makethe predictions look so wiggly? This wiggliness is presumably an indication that within a state, or maybe even across states, counties nearby can have large diferences, above and beyhond those that would be induced just by the prevalence of depression. So in a way, it might be that points that are side by side in the geographic space, and possibly with similar percentage voters for Trump, are actually in diferent plots (being located in different states).