6DoF VR Ventriloquist Effect - Supplementary Materials

Supplementary materials for the paper:

"Ventriloquist Effect in 6-Degrees-of-Freedom Virtual Reality with Object-Based Spatial Audio"
Bartłomiej Mróz · Paweł Perkowski · IEEE Transactions on Visualization and Computer Graphics, 2026 [major revision under review]

The repository state corresponding to the originally submitted manuscript is preserved under the tag tvcg-submission-v1. Renamed from ventriloquist-effect-in-6dof-vr-with-hoa during the revision.

This repository contains:

• Raw experimental data (localization responses + head trajectories)
• The reproducible statistical analysis pipeline
• Machine-readable analysis outputs, including the LaTeX value macros used by the manuscript

For methodology, interpretation, and results discussion, please refer to the main paper.

Repository Structure

.
├── data/
│   ├── LocalizationTestData_*.json   # Trial-level localization responses (31 participants)
│   └── OculusTrajectoryData_*.csv    # Head tracking trajectories (~9 Hz)
├── rdocs/
│   └── statistical_analysis.R        # Original analysis script (as submitted; see tag)
├── rscripts/revision/                # Revision analysis pipeline (current)
│   ├── build_analysis_df_revision.R  # Dataset construction incl. trajectory-derived measures
│   ├── refit_*.R                     # glmmTMB re-estimation of all Gamma GLMMs
│   ├── reviewer1_*.R, disparity_*.R  # Interaction, trial-number, nonlinearity, breakpoint analyses
│   ├── angular_* / verify_*.R        # Experienced angular disparity; independent verifications
│   ├── regenerate_tolerance_lookup.R # Developer tolerance table (paper Table I)
│   └── generate_revision_values.R    # Writes reported statistics into the manuscript as LaTeX macros
├── results/
│   └── analysis_report.md            # Original analysis output (as submitted)
├── results_revision/                 # Machine-readable outputs of the revision pipeline
└── README.md

Data

Participants: 31
Trials per participant: 24
Total observations: 744

Experimental variables:

• Sound type: 4 levels (Drum, Flute, Speech, Pink Noise)
• Audio-visual disparity: 15–70 cm (quasi-continuous, 56 distinct values)
• Spatial configuration: Azimuth and elevation of visual cue
• 6DoF head movement: Full exploratory movement allowed

Key measures:

• participantError_m: Unsigned localization error (meters)
• signedError_m: Bias toward (+) or away from (−) visual cue
• ventriloquistBias: Signed error as proportion of disparity
• Trajectory-derived (revision): experienced angular disparity, head-to-source distance, cue eccentricity, movement rate

Reproducing the Analysis

Prerequisites

install.packages(c("jsonlite", "tidyverse", "lme4", "lmerTest", "glmmTMB", "emmeans", "DHARMa", "mgcv"))

Run the Pipeline

Rscript rscripts/revision/build_analysis_df_revision.R   # builds results_revision/analysis_df_revision.rds
Rscript rscripts/revision/generate_revision_values.R     # regenerates the manuscript's reported values

Each script in rscripts/revision/ is self-contained and documents its inputs and outputs in a header comment; the verify_*.R scripts independently re-estimate the headline models. The per-sample trajectory intermediate (trajectory_samples_angular.rds, ~17 MB) is not tracked; build_analysis_df_revision.R regenerates it from the raw CSVs.

Note on estimation: the Gamma GLMMs are fitted with glmmTMB; the originally reported lme4::glmer Gamma fits did not converge reliably (degenerate variance-covariance matrices) and are retained only under the tvcg-submission-v1 tag for the record.

Citation

If you use this data or analysis code, please cite the main paper:

[BibTeX citation will be added upon publication]

License

This work is licensed under a Creative Commons Attribution 4.0 International License.

Contact

Bartłomiej Mróz · bartlomiej.mroz@pg.edu.pl · Department of Multimedia Systems, Gdańsk University of Technology · bmroz.eu