Changelog • rcicrdiagnostics

rcicrdiagnostics 0.2.1

Maintainer email updated to institutional address (m.j.barbosa.de.oliveira@tue.nl).
ORCID 0000-0002-6220-0695 added to author metadata in DESCRIPTION and .zenodo.json. Affiliation added to .zenodo.json (Eindhoven University of Technology).

?infoval and vignette §6.9.1 now state explicitly that the reference simulator’s uniform 50/50 ±1 response sampling is exact under the standard Brief-RC trial structure (equal numbers of oriented and inverted faces per trial: 6/6 in Brief-RC 12, 10/10 in Brief-RC 20), and call out that designs with unbalanced oriented/inverted splits will incur a small calibration drift.
Wording sweep: “canonical” replaced with more neutral synonyms (“original” for upstream-rcicr references; “typical”, “standard”, or “foundational” elsewhere) across user-facing documentation, NEWS, and tests.

New diagnose_infoval(): a guided six-step diagnostic for low or negative individual infoVal scores. Simulates a per-trial-count reference, runs a random-responder calibration check, computes per-producer and group-mean z (unmasked and optionally masked), tallies the per-producer distribution, and emits interpretation bullets. Supports both 2IFC and Brief-RC paradigms natively.
New infoval(): per-producer informational value with a reference distribution keyed on each producer’s actual trial count, closing the pool-size calibration gap in the original rcicr::computeInfoVal2IFC() for Brief-RC. Embeds the sampling-without-replacement fix from rcicrely v0.2.1. Exposes a with_replacement argument ("auto" / TRUE / FALSE) for controlling the across-trials reference sampling regime.
New face_mask(): parametric oval face-region masks with five sub-region variants (eyes, nose, mouth, upper, lower). Default geometry follows Schmitz, Rougier, & Yzerbyt (2024).
New load_face_mask(): reads PNG / JPEG mask images for users with hand-painted or webmorphR-produced masks.
run_diagnostics() gains face_mask and with_replacement arguments and runs diagnose_infoval() automatically when infoval_iter is supplied with the relevant noise source.
Brief-RC infoVal is now fully supported through diagnose_infoval() and the in-package infoval(). No rcicr install needed for Brief-RC; the path uses an in-package implementation of Schmitz’s genMask() algebra.

compute_infoval_summary() no longer fails on a per-participant headcount of z >= 1.96. Status now returns pass when the median individual z is positive, warn otherwise. Per the published record (Brinkman 2019: mean ~3-4 on 2IFC gender; Schmitz 2024: Brief-RC systematically below 1.96), individual infoVal magnitudes are paradigm- and target-dependent, so a fail status from a headcount alone misclassifies many compliant studies.
The data$per_participant$meaningful column produced by compute_infoval_summary() is renamed to above_threshold. Code that reads the old column name will break and needs the rename.

build_inputs_2ifc() (called from diagnose_infoval(method = "2ifc")) previously expected a stimuli 3D array inside the rcicr .RData, which does not exist. The 2IFC path now reconstructs per-trial noise patterns from stimuli_params and p via rcicr::generateNoiseImage(). The 2IFC diagnose_infoval() path was non-functional before this release.

Major vignette overhaul. Section 3 expanded into a full data-preparation reference (input format, example datasets for both pipelines, plain-language description of every object inside the rcicr .RData, noise matrix specification, base-image preparation pipeline via webmorphR, three ways to obtain face-region masks). New section 3.4.1 covers Brief-RC sampling regimes (without/with replacement) and their downstream effects. Section 6.9.1 rewritten with sourced empirical baselines from Brinkman 2019 and Schmitz 2024 replacing previously unsourced framing. New section 6.9.2 walks through diagnose_infoval().

Adds jpeg and png to Suggests for the optional load_face_mask() PNG/JPEG path. No new hard dependencies.