Pre-compile a linear mixed model template — precompile

Fits a lme4::lmer model (on each imputed dataset) using a dummy outcome variable. It produces a list of fitted model templates that can be updated with vertex outcome data, avoiding the overhead of re-specifying the random-effects structure for every vertex/surface measure.

Usage

precompile_model(
  formula,
  data_list,
  tmp_y,
  measure,
  weights = NULL,
  REML = TRUE,
  lmm_control = lme4::lmerControl(calc.derivs = FALSE),
  verbose
)

Arguments

formula: A two-sided formula object passed to lme4::lmer(). The left-hand side is replaced by a dummy outcome column named vw_<measure> before fitting.
data_list: A list of data.frame objects, typically constructed by imp2list() Each element must contain all variables referenced by formula.
tmp_y: A numeric vector used as the dummy outcome. Must have the same length as the number of rows in each element of data_list.
measure: A character string naming the brain surface measure (e.g. "thickness", "area"). Used to construct the temporary outcome column vw_<measure> injected into each imputed dataset.
weights: Either NULL (no observation weights), a numeric vector of weights of the same length as the rows in each imputed dataset, or a character string giving the name of a column in each dataset that contains the weights.
REML: Logical. Whether to use restricted maximum likelihood estimation. Passed directly to lme4::lmer(). Defaults to TRUE.
lmm_control: A lmerControl object produced by lme4::lmerControl(), used to fine-tune the optimiser behaviour. Defaults to lme4::lmerControl(calc.derivs = FALSE).
verbose: Logical. If TRUE, emits a cli progress step during model construction.

Value

A list of the same length as data_list, where each element is a fitted lmerMod object estimated on the corresponding imputed dataset with the dummy outcome. These objects are intended for use with lme4::refit() or equivalent downstream functions in the verywise pipeline.

Author

Serena Defina, 2026