Datasetpublic

boazkg/iDaVIE-particle-selection-vr-vs-desktop

VR vs desktop particle selection: user-study data (iDaVIE-p) Anonymised data for the within-subjects study reported in Particle Selection in Virtual Reality (Keren-Gil, Sivitilli, Marchetti, Gain, Marais; University of Cape Town / IDIA; submitted to Virtual Reality, Springer). Thirty-four participants selected 90 target galaxies out of the 16,816-galaxy 2MRS Galaxy Group Catalogue (the 2MASS Redshift Survey galaxies that belong to identified groups), once in VR (Meta Quest 2… See the full description on the dataset page: https://huggingface.co/datasets/boazkg/iDaVIE-particle-selection-vr-vs-desktop.

sourceHugging Facecc-by-4.0updated 11h agoView on Hugging Face

0likes65downloads

Dataset Card

VR vs desktop particle selection: user-study data (iDaVIE-p)

Anonymised data for the within-subjects study reported in Particle Selection in Virtual Reality (Keren-Gil, Sivitilli, Marchetti, Gain, Marais; University of Cape Town / IDIA; submitted to Virtual Reality, Springer). Thirty-four participants selected 90 target galaxies out of the 16,816-galaxy 2MRS Galaxy Group Catalogue (the 2MASS Redshift Survey galaxies that belong to identified groups), once in VR (Meta Quest 2, spherical brush) and once on a desktop (mouse/keyboard, single-click and 3D selection box), and answered the SUS, NASA-TLX and Flow State Scale after each task.

Outcome measures are the per-session summaries recorded by the software (true/false positives, false negatives, completion time) and the questionnaire item responses; per-galaxy selection logs are not part of this release. Everything in the paper's Results section can be regenerated from this folder:

bash

python analysis/reproduce_stats.py   # needs numpy + scipy; prints every table and statistic
python analysis/make_figures.py      # needs numpy + matplotlib; redraws the two box-plot figures

analysis/REPRODUCE_OUTPUT.txt is the output of that script at release time.

Study design (verified against the raw records)

	Group 1 (n = 17)	Group 2 (n = 17)
First task	VR, target set A	Desktop, target set A
Second task	Desktop, target set B	VR, target set B

Platform order is counterbalanced across groups (17/17); each target set was used on each platform by 17 participants; set A was always the first task (stated as a limitation in the paper). Order was cross-checked against the questionnaire submission timestamps for all 34 participants. All participants had a 15-minute VR playground on a mock dataset before the first task; each task was capped at 600 s.

Layout

data/          per-participant outcomes and all questionnaire item responses
codebook/      item wording, scales, scoring rules; registration-form coding
target_sets/   the 2MRS catalogue with the 90 targets of set A and set B flagged
analysis/      scripts that reproduce every statistic and both box-plot figures in the paper

File	Contents
`data/participants.csv`	One row per participant (P01–P34, all 34 present): design cell, demographics (coded), and per-platform outcomes: TP/FP/FN, accuracy, completion time, SUS, NASA-TLX composite, FSS total, six raw NASA-TLX subscales, nine FSS subscales.
`data/questionnaire_responses.csv`	Long format, one row per item response: `participant_id, platform, order_position (1 = first task), item_id, response`. 68 questionnaires × 52 items.
`codebook/questionnaire_items.csv`	Codebook: item id, instrument, subscale, exact item wording, response scale, scoring rule.
`codebook/registration_form.md`	The pre-experiment form and how each field was coded/banded.
`target_sets/set_A_…tbl`, `set_B_…tbl`	The full 2MRS group-galaxy catalogue used in the study (16,816 rows) with a `test` column flagging the 90 targets of each set (80 clustered + 10 scattered; the two sets are disjoint).
`target_sets/rendering_config_2MRS.json`	iDaVIE-p rendering/mapping configuration used for the catalogue (colour map driven by distance, uniform opacity 0.5).
`analysis/reproduce_stats.py`, `analysis/REPRODUCE_OUTPUT.txt`	Recomputation of all reported statistics.
`analysis/make_figures.py`, `analysis/figures/`	The paper's two box-plot figures (accuracy and completion time; SUS, NASA-TLX and FSS), redrawn from `data/participants.csv`.

Scoring conventions

—Accuracy = TP / (TP + FP + FN) over the 90 targets (true negatives excluded).
—Completion time in seconds, capped at 600.
—SUS: standard scoring (odd items: response − 1; even items: 5 − response; sum × 2.5).
—NASA-TLX: six raw subscales on 0–100 (steps of 5), unweighted; composite = mean of the six. (The study's original scoring sheet rounded the composite up to an integer; this release and the revised paper use the exact mean.)
—FSS (Flow State Scale; Jackson & Marsh 1996; 36 items, 1–5): total = mean item score × 20 (0–100); subscale k = mean of its four items (k, k+9, k+18, k+27 in questionnaire order) × 20. (The original scoring sheet used a wrong item mapping for the subscales; the values here are the corrected ones used in the revised paper.)
—Statistics: Wilcoxon signed-rank (two-sided) for paired comparisons; Cohen's d with pooled SD, Desktop − VR; Holm–Bonferroni within each family of tests; sample SDs.

Anonymisation

Participant IDs are random and unrelated to session order. Names, contact details, dates of birth and session dates/times are not included; age is banded; department and academic level are as recorded on the registration form (coarse categories). The only temporal information retained is minutes_between_questionnaires (task-1 to task-2 questionnaire gap) and order_position.

Ethics and licence

The study was approved by the University of Cape Town Faculty of Science research ethics committee; participants gave written informed consent, including to the use of anonymised data for research and publication. Data are released under CC BY 4.0. The 2MRS catalogue columns derive from Huchra et al. (2012) and the 2MRS group catalogue of Lambert et al. (2020); please cite those works when reusing the catalogue.

Citation

Keren-Gil, B., Sivitilli, A., Marchetti, L., Gain, J., Marais, P. (2026). Particle Selection in Virtual Reality: VR vs desktop user-study data (iDaVIE-p) [Data set]. Hugging Face. https://huggingface.co/datasets/boazkg/iDaVIE-particle-selection-vr-vs-desktop