Stimulus generalization

In gp3ml: Governance-First Predictive Modelling for 'Gazepoint' Research

Stimulus-grouped generalization workflow

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

data <- simulate_gazepoint_governed_data(18L, 9L, 1L, seed = 2501L)
predictors <- c("tracking_ratio", "fixation_duration", "gaze_dispersion")
task <- create_gazepoint_synthetic_task(data, "recording_quality", "new_stimuli")
manifest <- create_gazepoint_synthetic_manifest(task$outcome, predictors)
folds <- create_gazepoint_group_folds(
  data, task$outcome, predictors, manifest,
  "new_stimuli", "participant_id", "trial_id", "stimulus_id",
  v = 3L, repeats = 2L, seed = 2501L
)
diagnose_gazepoint_group_folds(folds)
evaluation <- evaluate_gazepoint_group_folds(
  folds, task, predictors, "glm", seed = 2501L
)
summarize_gazepoint_resample_uncertainty(evaluation, unit = "fold")

Stimulus-grouped assessment estimates generalization to held-out stimuli only. It does not establish participant generalization.

gp3ml documentation built on Aug. 23, 2026, 5:11 p.m.