Data, code and figures for "What a Five-Trial Pilot Can and Cannot Establish: Design and Evidential Characterisation of a Marker-Based Augmented Reality Application for Indonesian Traditional Musical Instruments"

Replication package for a marker-based mobile augmented reality application presenting five Indonesian traditional musical instruments (Bonang, Gambang, Angklung, Sampe, Tifa), and for the statistical characterisation of its pilot evaluation data. This is a small dataset, deposited deliberately. It contains five detection-latency measurements and four illumination conditions — the entire empirical content of a student pilot study. The associated manuscript does not present these as a validation of the system; it asks what a dataset of this size licenses, and finds the answer strongly asymmetric. The latency data support a firm inferential claim: a one-sided t-test rejects the hypothesis that the true mean reaches the one-second interaction threshold (t(4) = -11.65, p = 1.6e-4), establishing sub-second responsiveness at n = 5. The reliability data support almost nothing: four consecutive successes bound detection probability only above 0.473 at 95% confidence, and certifying 90% reliability would require 29 consecutive successes. Contents: four CSV data files with a full data dictionary; Python code reproducing every number and figure in the manuscript in under 30 seconds; the archived output of the reference run; five publication-quality figures; and the five unmodified original screenshots. Known limitations are documented in data/data_dictionary.md and include two that weaken the study's own conclusions: latency was timed with a manual stopwatch whose error is of the same order as the measured quantity, and the illumination design used an emissive (LCD) marker medium, which makes it incapable of falsification and non-comparable with printed-marker studies. No human-participant data are present.