workflow for `zi_predict`

[joshyam-k]

• prediction will now happen outside of both future_map and map respectively and thus any population data does not need to get passed to either

predictions options:

• return full models from each fitting and then just use predict (probably slow and memory intensive)
• return model parameters and figure out how to use matrix multiplication to do all of the prediction at once (outsource to c++ if needed) (hopefully the better option)

[joshyam-k]

saving this snippet of code that will help with some of the new prediction work

predict_test <- function(lmmod, glmmod, .data, domain_level) {
  lookup <- setNames(
    lmmod@u,
    model.frame(lmmod)[[domain_level]] |>
      unique() |>
      as.vector()
  )

  .data |>
    mutate(!!domain_level := lookup[.data[[domain_level]]])
}

[joshyam-k]

closing