Extracting data used to make a smooth plot in mgcv

Question

This thread from a couple of years ago describes how to extract data used to plot the smooth components of a fitted gam model. It works, but only when there is one smooth v

Answer 1

Gavin gave a great answer, but I wanted to provide one in terms of the original referenced post (as I just spent a good amount of time figuring out how that worked :).

I used the code directly from https://stats.stackexchange.com/questions/7795/how-to-obtain-the-values-used-in-plot-gam-in-mgcv and also found that I only got the last model returned. The reason for that is because of where the trace code snippet is being placed in the mgcv::plot.gam function. You need to make sure that the code is placed inside a for loop that iterates over m, and you control that by the at argument.

The following trace worked great for my version of mgcv:::plot.gam

plotData <<- list()
trace(mgcv:::plot.gam, at=list(c(26,3,4,3)), 
quote({
       plotData[[i]] <<- pd[[i]]
  })
)

It inserts the trace call right after this chunk in the mgcv:::plot.gam function:

if (m > 0) 
    for (i in 1:m) if (pd[[i]]$plot.me && (is.null(select) || 
        i == select)) {

and now the elements of plotData will correspond to the different variables plotted. Two functions I found very helpful for figuring out the right place to insert this trace call were

edit(mgcv:::plot.gam)
as.list(body(mgcv::::plot.gam))