Run leave-one-unit model sensitivity analysis — run_gazepoint_model_leave_one

Refit the same model repeatedly while removing one participant, item, stimulus, trial, or other analysis unit at a time. The helper compares leave-one-out estimates with the full-data model to assess whether a key effect is driven by a single unit.

Usage

run_gazepoint_model_leave_one_out(
  data,
  unit_col,
  fit_function,
  extract_function = NULL,
  effect_terms = NULL,
  min_rows = 2L,
  keep_models = FALSE,
  name = "gazepoint_model_leave_one_out"
)

Arguments

data: A data frame used for model fitting.
unit_col: Column identifying the unit to leave out, for example subject, participant, item, stimulus, or trial.
fit_function: Function that takes one data frame argument and returns a fitted model.
extract_function: Optional function that takes a fitted model and returns a data frame of effects. If NULL, a default coefficient extractor is used for common model objects.
effect_terms: Optional character vector of terms/effects to retain in the sensitivity summary.
min_rows: Minimum number of rows required after leaving one unit out.
keep_models: Logical. If TRUE, keep the full model and refitted models.
name: Character label stored in the returned object.

Value

A list with class gp3_model_leave_one_out_sensitivity.

Details

This is a generic robustness wrapper. It can be used with linear models, GLMs, mixed models, GAMMs, GCA models, AOI GLMMs, pupil LMMs, or any custom model as long as a fitting function is supplied.