Probe code, reachability and sensitivity sweeps, figure code and manuscript source for a reproducibility audit that records, rung by rung, where reconstruction of a published acoustic-localisation comparison stops on one ordinary machine and date. Sensitivity panels are descriptive; the stopping rung is not a universal method verdict.

Archived at 10.5281/zenodo.22647014.

Repository: https://github.com/PeterPonyu/comparison-reachability-audit

• paper/tex/ — manuscript source
• paper/figs/ — the R code that draws the figures and writes the printed numbers
• paper/evidence/ — a file list with SHA-256 hashes
• data/ — the 18 data files named in that list

This archive leaves out one extra file named in the paper's evidence list. The paper does not take any number from it.

• A private working note. The one count the paper uses from it -- how large the missing corpus is -- is also written in the inventory file included here.

bash build.sh

The build checks every data file against its hash and stops if a file has changed. Figures and printed numbers are generated from those files, not typed in by hand.

Requires python3, Rscript with digest, ggplot2, jsonlite, patchwork and systemfonts, and a TeX distribution with latexmk.

Working draft. Not submitted to any venue.

Code: MIT (LICENSE). Manuscript text, figures and recorded result data: CC BY 4.0 (LICENSE-CONTENT).