irt_validation_spec: Specify a validation programme for a process-IRT model

In eyeprocess: Harmonize Eye-Tracking, Pupillometry, Biometrics, and Psychometric Process Data

Specify a validation programme for a process-IRT model

Description

Creates an evidence contract rather than executing a particular estimator. The object can be passed to stress-test and evidence-grading helpers.

Usage

irt_validation_spec(
  model_id,
  replications = 250L,
  parameters = NULL,
  metrics = c("bias", "rmse", "coverage", "interval_width", "convergence"),
  grouped_validation = c("device", "session", "site"),
  preprocessing_variants = NULL,
  misspecification_scenarios = NULL,
  thresholds = list(max_abs_bias = 0.1, max_rmse = 0.3, min_coverage = 0.9,
    max_failure_rate = 0.05, min_external_folds = 2L),
  seed = 20260808L,
  notes = NULL
)

Arguments

model_id	Stable model identifier.
replications	Number of simulation replications planned.
parameters	Parameter families expected to be recovered.
metrics	Recovery metrics to require.
grouped_validation	Grouping variables for transport validation.
preprocessing_variants	Optional named preprocessing variants.
misspecification_scenarios	Optional named scenarios.
thresholds	Named evidence thresholds.
seed	Reproducibility seed.
notes	Free-text scientific notes.

Value

An object of class "eye_irt_validation_spec", stored as a named list, with components "model_id", "replications", "parameters", "metrics", "grouped_validation", "preprocessing_variants", "misspecification_scenarios", "thresholds", "seed", "notes", "contract_version", "created_with". It contains specify a validation programme for a process-IRT model and associated metadata or diagnostics needed to interpret the result.

eyeprocess documentation built on Sept. 28, 2026, 5:08 p.m.