Publications

Preprints and work in progress

• Goodness-of-fit for distributions on metric spaces

  Serrano, D., García-Portugués, E. and Van Keilegom, I. (2026). arXiv:2609.38158.

  A goodness-of-fit framework for distributions on separable metric spaces, built on distance profiles, which characterise a distribution under suitable identifiability conditions. It covers simple and composite null hypotheses, incorporating parameter estimation through a Bahadur-type expansion, and gives the asymptotic null distributions of the test statistics, their consistency under fixed and local alternatives, and multiplier bootstrap procedures. Illustrated on the sphere, the hyperboloid and the simplex.

  arXiv PDF DOI
• Out-of-bag prediction balls for random forests in metric spaces

  Serrano, D. and García-Portugués, E. (2025). arXiv:2510.04299.

  Confidence regions around predictions from any bagged regression algorithm with metric-space-valued responses, built from the out-of-bag observations generated in a single forest training. Asymptotic guarantees for four coverage types, and smaller, more stable balls than split-conformal methods on the sphere, the hyperboloid and the space of positive definite matrices. Applied to solar dynamics.

  arXiv PDF DOI

Book chapters and proceedings

• Prediction regions for functional-valued random forests

  Serrano, D. and García-Portugués, E. (2025). In G. Aneiros, E. G. Bongiorno, A. Goia and M. Hušková (eds.), New Trends in Functional Statistics and Related Fields, Contributions to Statistics, pp. 475–482. Springer, Cham.

  Prediction regions for random forests with functional output, based on a metric specification and on Fréchet regression, using out-of-bag observations to quantify predictive uncertainty. Proceedings of IWFOS 2025, Novara, Italy.

  PDF DOI