Modify a mixed-effects formula in R

Question

I have a formula:

my_formula <- a ~ b + (1|d) + e

I want to programmatically remove the e variable from this formula.

What I've done for fixed-effects only formulae is:

drop_es <- function(f) {
    e_idxs <- grep(pattern = 'e',
                   x = all.vars(f[[3]]))

    stats::formula(stats::drop.terms(termobj = stats::terms(f),
                                     dropx = e_idxs,
                                     keep.response = TRUE))
}

This works on fixed-effects only formula:

drop_es(f = a ~ b + e)

a ~ b

But on a formula with a random effect, it drops the parentheses:

drop_es(f = a ~ b + (1|d) + e)

a ~ b + 1 | d

This is a really important distinction -- without the parentheses, it is interpreted as:

a ~ b|d

How can I remove the e terms without losing the parentheses?

Answer 1

The update function exists to modify formulas. use

update(my_formula, ~.-e)
# a ~ b + (1 | d)

The -e means drop the "e" term from the formula while . means keep everything else.