Reviewer-facing QC sequence
This article shows a compact quality-control workflow for Gazepoint-derived biometric and eye-tracking exports. The workflow is conservative: it validates dataset layout, metadata, missingness, signal quality, exclusion decisions, and reproducibility outputs. It does not infer clinical, affective, emotional, diagnostic, or psychological states.
Dataset layout
layout_audit <- check_gazepoint_bids(
root = "path/to/gazepoint_dataset"
)
layout_auditcheck_gazepoint_bids() is a lightweight
Gazepoint-oriented layout audit. It is not a full BIDS validator and
does not convert data.
Metadata and required columns
metadata_audit <- validate_gazepoint_metadata(
data = gazepoint_data,
required_columns = c("participant_id", "time", "trial")
)
metadata_auditMissingness
missingness_summary <- summarize_gazepoint_missingness(gazepoint_data)
missingness_summary
plot_gazepoint_missingness(gazepoint_data)Missingness summaries and plots are descriptive QC tools. They help identify sparse recordings, dropouts, and participant-level recording problems.
Pupil and signal artifacts
pupil_qc <- detect_gazepoint_blinks(
gazepoint_data,
pupil_cols = c("pupil_left", "pupil_right")
)
nonwear_qc <- detect_gazepoint_nonwear(
gazepoint_data,
signal_cols = c("eda", "ppg", "ecg")
)Masking, interpolation, filtering, and beat correction should be reported as transparent preprocessing steps.
Signal quality and rule-based flags
signal_quality <- compute_gazepoint_signal_quality(
gazepoint_data,
signal_cols = c("eda", "ppg", "ecg")
)
quality_summary <- summarize_gazepoint_signal_quality(signal_quality)
quality_flags <- classify_gazepoint_signal_quality(
signal_quality,
rules = list(
missingness_rate = 0.20,
quality_index = 0.50
)
)Rule-based quality flags support auditability. They are not clinical, psychological, or affective interpretations.
Exclusion recommendations
exclusion_recommendations <- recommend_gazepoint_biometric_exclusions(
quality_flags
)
exclusion_recommendationsExclusion helpers summarize evidence for review. They support analyst judgment but do not replace it.
Static dashboard
dashboard <- pipeline_comparison_dashboard(
quality_flags,
participant_col = "participant_id",
session_col = "session",
missingness_col = "missingness_rate",
quality_col = "quality_index",
qc_status_col = "qc_status",
failed_rules_col = "failed_rules",
excluded_col = "excluded",
notes_col = "audit_notes"
)
dashboardpipeline_comparison_dashboard() creates a static
reviewer-facing summary. It is not a Shiny app, GUI dashboard, model, or
interpretation layer.
Reproducibility outputs
manifest <- create_gazepoint_analysis_manifest(
input_files = list.files("path/to/gazepoint_dataset", recursive = TRUE),
parameters = list(
missingness_threshold = 0.20,
quality_threshold = 0.50
)
)
dictionary <- create_gazepoint_dictionary(gazepoint_data)Recommended reporting language
Gazepoint-derived exports were processed using conservative, rule-based quality-control checks. Dataset layout, metadata fields, missingness, signal quality, artifact flags, and exclusion recommendations were summarized before analysis. The workflow produced audit-ready summaries and reproducibility metadata, but did not infer clinical, affective, emotional, or diagnostic states from biometric signals.