Simultaneous participant and stimulus generalization
In gp3ml: Governance-First Predictive Modelling for 'Gazepoint' Research

Participant-and-stimulus generalization workflow

knitr::opts_chunk$set(collapse = TRUE, comment = "#>")
options(gp3ml.reproducible_examples = TRUE)
library(gp3ml)

data <- simulate_gazepoint_governed_data(18L, 6L, 1L, seed = 2601L)
predictors <- c("tracking_ratio", "blink_rate", "gaze_dispersion")
task <- create_gazepoint_synthetic_task(
  data, "recording_quality", "new_participants_and_new_stimuli"
)
manifest <- create_gazepoint_synthetic_manifest(task$outcome, predictors)
folds <- create_gazepoint_group_folds(
  data, task$outcome, predictors, manifest,
  task$generalization_target,
  "participant_id", "trial_id", "stimulus_id",
  v = 3L, repeats = 1L, seed = 2601L
)
folds$fold_summary
evaluation <- evaluate_gazepoint_group_folds(
  folds, task, predictors, "glm", seed = 2601L
)
evaluation$fold_status
head(evaluation$excluded)

Cross-block rows that belong to only one held-out grouping dimension remain explicitly excluded. They are not silently reassigned to analysis or assessment.