RegressionFormulae.jl

Base.:/ — Method

group / term

Generate predictors for term within each level of group. Implemented as group + fulldummy(group) & term.

Base.:^ — Method

(term1, term2, ...) ^ n

Generate all interactions of terms up to order $n$.

Warning

Embedded InteractionTerms (i.e. (a + b + c & d)^2) are not currently supported and result in an error.

fulldummy(term::CategoricalTerm)

Assign "contrasts" that include all indicator columns (dummy variables) regardless of whether an intercept column is present.

Note that this generally results in an overparameterized model, but may be valid in regularized models.