Closed gavinsimpson closed 2 years ago
As an extreme example, see this for the Swiss extreme rainfall example data set used in the second edition of Simon Wood's book:
require("gamair")
library("gratia")
library("mgcv")
data(swer, package = "gamair")
m <- gam(list(exra ~ s(nao) + s(elevation) + climate.region + s(N, E),
~ s(year) + s(elevation) + climate.region + s(N, E),
~ climate.region),
family = gevlss(),
data = swer)
draw(m)
with the draw(m)
resulting in
Where the smooths for the scale parameter of the GEV come before those for the location parameter, and are in the wrong order within those parameters!
I think something must be going on to reorder the smooth column somewhere in all the group_by, mapping, nesting/unnesting, binding, etc that happens in
draw.gam()
An easy solution should be to use
smooths()
to reorder the column into the order that the smooths entered the model.