Now predict_gam() uses the lpmatrix predictions which means that when excluding terms the user will have to set a value for the excluded variable.
This is especially relevant when excluding (random) factor smooths: the user should pick a value for the random variable in the smooth to avoid the same predictions being output for all levels in the random variable (see the Get started vignette for examples).
Note that older code will still work but will produce duplicated predictions.
predict_gam() uses the lpmatrix predictions. It is now possible to exclude any terms, smooth or parametric.get_difference() returns difference between two smooths.
plot.tidygam.diff method to plot difference smooth.
Data gest and struct.
separate and sep_by arguments in predict_gam() allow the user to separate variables in the model that were created with interaction().
Vignette get-started.Rmd.
Error when predicting bivariate smooths (s/te/ti(), fs/re basis functions) where only the first variable was returned internally.
Error when plotting difference smooth with no significant difference (closes #5).
Error where get_difference() did not work when excluding random smooths (closes #8).
First minor release of the package.
Added a NEWS.md file to track changes to the package.