Repository for the paper: “Laplace Approximations for Mixed-Effects and Gaussian Process Quantile Regression”.

This repository accompanies our paper on Laplace-based inference for (i) Gaussian-process (GP) models and (ii) mixed models. The codebase is organized to support fully reproducible experiments, including synthetic-data generation, method implementations in both Python and R, and scripts to run experiments and reproduce paper figures.

Branch overview

• GP branch: Gaussian-process (GP) experiments and GP-based methods.
• mixed-model branch: mixed-model experiments and mixed-model methods.

If you are looking for GP-related code, use the GP branch.
If you are looking for mixed-model code, use the mixed-model branch.

---

You are on branch: GP

This branch contains the GP experiments and (where relevant) the notebooks used to generate paper plots.

Repository structure

• R/
  R implementations of selected methods and supporting routines used in the experiments.

• src/
  Python source code: model implementations and utilities (e.g., data generation helpers, plotting utilities, evaluation metrics).

• scripts/
  Experiment entry points and orchestration scripts. Typical tasks include:

  • generating datasets,
  • running Python methods,
  • merging results produced by Python and R,
  • producing summary tables/plots for the paper.

• configs/
  Experiment configuration files (e.g., YAML) defining simulation settings, hyperparameters and run grids.

• notebooks/ Notebooks used to reproduce selected plots/figures for the paper.

• data/real_data/
  Real dataset(s) used in the empirical experiments (tracked).
  See the folder contents for file-level details.

Reproducing experiments (high level)

1. Set experiment parameters in configs/.
2. Run the appropriate entry script(s) from scripts/ to generate data and fit models.
3. Use the analysis/aggregation scripts in scripts/ to merge outputs (Python + R) and produce tables/figures.

Environment. The experiments were run with Python 3.13.2 and R 4.4.0.

Notes on outputs

Intermediate outputs and large result artifacts are typically written to results/ (often not tracked in Git to keep the repository lightweight).

Citation

If you use this code, please cite the corresponding paper.