Antonio Aguirre Bayesian Forecasting | Quantile Modeling | Research Software

Research

I develop Bayesian time-series methods for probabilistic forecasting, with an emphasis on quantile modeling, uncertainty quantification, scalable inference, and reusable research software.

Much of this work is motivated by hydrological and environmental forecasting, where forecasts need to be calibrated, interpretable, computationally practical, and easy to update as new data arrive.

Forecasting

Forecast Correction and Synthesis

Bayesian quantile methods for correcting river-flow forecasts, combining forecast sources, and evaluating predictive distributions with proper scoring rules.

Nonlinear Time Series

Q-DESN Quantile Forecasting

Bayesian quantile readouts for fixed Deep Echo State Network features, with regularized readouts, exAL working likelihoods, MCMC, and variational approximations.

State Space Models

Dynamic Quantile Models

Flexible dynamic quantile linear models with trend, seasonal, regression, transfer-function, forecasting, diagnostics, and posterior synthesis components.

Computation

Scalable Inference and Diagnostics

Variational Bayes, Sequential Monte Carlo, simulation diagnostics, and reproducible workflows for models that need to run repeatedly and be checked carefully.

Selected Papers & Software

  • Submitted / CRAN
    Aguirre, A., Barata, R., Prado, R., Sansó, B. exdqlm: An R Package for Estimation and Analysis of Flexible Dynamic Quantile Linear Models. Manuscript submitted to the Journal of Statistical Software; companion package available on CRAN.
  • Working paper
    Aguirre, A., Prado, R., Sansó, B. Bayesian Quantile Readouts for Fixed Deep Echo State Networks. Current work on Q-DESN quantile forecasting.
  • In revision
    Aguirre, A., Sansó, B., Prado, R. A Bayesian Quantile-Based Correction and Synthesis of River Flow Forecasts. Manuscript in revision.
  • Published
    Aguirre, A., Lobato, I.N. (2024). Evidence of Non-Fundamentalness in OECD Capital Stocks. Empirical Economics. DOI.

Code and Reproducibility

Selected package code, manuscript-support scripts, and data-processing workflows are collected on the Software page. I keep that page selective so each public example has a clear purpose and enough context to be useful.