Persistent false species detections and bias in repeated visit occupancy estimates — manuscript, code and simulation data
Description
Research manuscript and accompanying reproducibility materials for a synthetic repeated-visit occupancy study. The manuscript examines bias caused by persistent false species detections. The executed numerical experiment uses 1,000 sites, 200 replicates per condition, occupancy 0.3, detection probability 0.6, false-positive probability 0.02, five visit counts and three correlation settings; 3,000 replicate rows are supplied. No wildlife observations, animal experiments or trained classifier performance are claimed. Files: manuscript.pdf is the full research manuscript; run_s6.py is the original numerical generator; s6_replicates.csv contains replicate-level estimates; summary.json contains aggregated results; README.txt documents variables, assumptions, reproduction and provenance. AI disclosure: OpenAI Codex assisted with study formulation, implementation, execution, automated checking and manuscript drafting. Numbers were obtained by executing code, not by asking an LLM to invent results. Automated reruns are not independent human validation. The depositor has confirmed the data are real and reliable. This deposit is a preprint/research-material record and does not imply peer-reviewed journal acceptance.
Files
Steps to reproduce
Use the Python/NumPy versions documented in the supplied environment file. Run python run_s6.py from a working directory. The code preserves child stream 5 of six SeedSequence streams from seed 20261004. Compare generated s6_replicates.csv and summary.json with the archived results. See README.txt for column definitions and study limitations.