I am looking for a two-dimensional version of the smooth.spline function in R, but cannot find one.
Following a suggestion of Cosma Shalizi (https://www.stat.cmu.edu/%5C~cshalizi/TALR/TALR.pdf, page 368) I am trying to build an iterative weighted least square fit, but for a regression with two predictors (instead of one predictor, as described in the text).
The central idea of the method is to weigh the predictor(s) by the associated variance, which is estimated from a spline fit to the log of the squared residuals.
For one predictor, R's `smooth.spline` function is used to obtain that fit. I have been looking for a similar R function that would work for two (or even more) dimensions, but could not find any. Any suggestion would be welcome!