Extrapolation of extreme covariates in generalized additive regression using extreme value theory
Abstract
Predictions with covariates in the tails of the covariate distribution often lack accuracy and are highly uncertain but may be of critical importance in many applications. New covariates may even lie beyond the range of training data. The problem can be particularly critical in environmental contexts, for example with climate-change scenarios, where new covariates represent future, more extreme climate conditions. We here propose novel methods to improve generalized additive models (GAMs) with focus on extreme covariates and covariate extrapolation. Adopting a random-design setting, we continuously integrate GAMs for the bulk of covariate distributions with models motivated by multivariate extreme value theory for high covariate values. We consider continuous responses but develop also a new approach for binary responses by assuming a continuous latent response variable. Our framework imposes a specific structure for large values of the covariates motivated by extreme value theory: by combining a transformation to a specific marginal scale with an appropriate link function, the continuous response variable depends linearly on the covariates. In an application to occurrences and sizes of large wildfires in Europe for the period 2008-2023, we explore how the new method can improve predictions, especially during extreme conditions, using environmental and meteorological covariates.
Related
- Doubly robust target inference for generalized linear regression with completely missing covariates
- Kernel Density Regression Estimation with Covariates on Riemannian Manifolds
- Byzantine-Robust Sparse Regression with Missing Covariates
- Augmented transfer regression learning for completely missing covariates
- BGM-IV: AI-Powered Bayesian Generative Modeling for Instrumental Variable Regression with High-Dimensional Covariates
BibTeX
@misc{carcaiso2026extrapolation,
title = {{Extrapolation of extreme covariates in generalized additive regression using extreme value theory}},
author = {Viviana Carcaiso and Sebastian Engelke and Juliette Legrand and Thomas Opitz},
howpublished = {HAL},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2607.08921},
eprint = {2607.08921},
archivePrefix = {arXiv},
url = {https://hal.science/hal-05797660v1},
note = {CC BY-NC-SA 4.0},
}
RIS
TY - UNPB
TI - Extrapolation of extreme covariates in generalized additive regression using extreme value theory
AU - Viviana Carcaiso
AU - Sebastian Engelke
AU - Juliette Legrand
AU - Thomas Opitz
PY - 2026
DA - 2026/09/14
PB - HAL
DO - 10.48550/arxiv.2607.08921
UR - https://hal.science/hal-05797660v1
AB - Predictions with covariates in the tails of the covariate distribution often lack accuracy and are highly uncertain but may be of critical importance in many applications. New covariates may even lie beyond the range of training data. The problem can be particularly critical in environmental contexts, for example with climate-change scenarios, where new covariates represent future, more extreme climate conditions. We here propose novel methods to improve generalized additive models (GAMs) with focus on extreme covariates and covariate extrapolation. Adopting a random-design setting, we continuously integrate GAMs for the bulk of covariate distributions with models motivated by multivariate extreme value theory for high covariate values. We consider continuous responses but develop also a new approach for binary responses by assuming a continuous latent response variable. Our framework imposes a specific structure for large values of the covariates motivated by extreme value theory: by combining a transformation to a specific marginal scale with an appropriate link function, the continuous response variable depends linearly on the covariates. In an application to occurrences and sizes of large wildfires in Europe for the period 2008-2023, we explore how the new method can improve predictions, especially during extreme conditions, using environmental and meteorological covariates.
ER -