Residual plots for gam checking

[nhcooper123]

Remind us of what the residual plots should be for checking the gam

[dill]

Okay, so we want something that looks at the per-covariate residuals over bins of the residual values. Something like...

load("Model_for_Dave.Rdata")
sst_bins <- cut(all_strandings$Max_SST, c(14,15,16,17,18))
sst_resid_data <- data.frame(SST=sst_bins, resids=residuals(All_strand))
boxplot(resids~SST, data=sst_resid_data)

which looks like this:

the variation in each boxplot should be pretty consistent if the model is good.