Statistical Modeling for Combinatorial Response Data Reproducibility Repository

This repository contains code and data to reproduce the results in the paper.

All commands must be run from the repository root.

Generated outputs are written to:

results/

First time? Set up R and Python dependencies first — see INSTALL.md.

---

Quick Start

1️⃣ Sanity check

make smoke
make probit_smoke

Runs a lightweight simulation and data analysis to verify everything works.

---

2️⃣ Main simulation study (MH-within-Gibbs)

make full

• The default setting does not run the full grid of simulation settings in the paper, to save time. To run the sampler with custom settings, change the corresponding environment variables first. See §6 in REPRODUCIBILITY.md for details.

• Note: scripts/run_full.sh defaults to JASA_METHODS="exponential halfgaussian", so make full produces results for both kernels in one pass. The main paper uses only the exponential kernel; the half-Gaussian results feed into Supplementary Table 3. To run only the main-paper kernel:

  JASA_METHODS=exponential make full

• Outputs saved under:

  results/runs/mh_within_gibbs/
  

---

3️⃣ Kernel comparison (supplementary)

make kernel_compare

• Runs both:

  • exponential
  • half_gaussian

• Outputs saved under:

  results/runs/mh_within_gibbs/kernel_compare/
  

---

4️⃣ Data analysis (waterfowl matching)

Full model:

make duck_full

Reduced model:

make duck_reduced

• Outputs saved under:

  results/data_analysis/<RUN_TAG>/
  

---

⚙️ Custom Settings

All scripts support environment-variable overrides. For more details, see §6 in REPRODUCIBILITY.md.

Example:

JASA_N_ITER=20000 make full

JASA_METHODS=exponential,half_gaussian make full

---

📄 Full Documentation

See:

REPRODUCIBILITY.md

for detailed mapping between paper sections and scripts.