Why does including latitude and longitude in a GAM account for spatial autocorrelation?
This thread is about how applying a smoother to longitude and latitude is a respectable method to deal with spatial autocorrelation.
I have an old-school professor who thinks this is nonsense, and that the only viable way is to model the correlation using a covariance structure imposed on residuals $\epsilon$.
I told him about the cross validated thread above, and he didn't took a look at it, claiming that there's a lot of wrong things said on forums like these.
So, now I am just looking for a better source: a book, an article, anything, which makes a good case for above claim. Any help?