EUVIMEDOpdag EUVIMED
Crossref

A comparison of some conformal quantile regression methods

Matteo Sesia, Emmanuel J. Candès

Stat · 2020-01 · 9 (1)

DOI: 10.1002/sta4.261 ↗

At a glance

Journal article in Stat, published online on 2020-02-19. Written by Matteo Sesia and 1 co-authors. Cited 62 times so far (Crossref); 27 references. Funded by: Army Research Office, Division of Mathematical Sciences.

AbstractBibliografiske oplysningerCiteRelated studies

Abstract (original language)

We compare two recent methods that combine conformal inference with quantile regression to produce locally adaptive and marginally valid prediction intervals under sample exchangeability (Romano, Patterson, & Candès, 2019, arXiv:1905.03222; Kivaranovic, Johnson, & Leeb, 2019, arXiv:1905.10634). First, we prove that these two approaches are asymptotically efficient in large samples, under some additional assumptions. Then we compare them empirically on simulated and real data. Our results demonstrate that the method of Romano et al. typically yields tighter prediction intervals in finite samples. Finally, we discuss how to tune these procedures by fixing the relative proportions of observations used for training and conformalization. Our empirical results suggest that using between 70% and 90% of the data for training often achieves a good balance between minimizing the average width of the predictions intervals and the variability in their practical coverage.

Funding

  • Army Research Office
  • Division of Mathematical Sciences

Cite

APA

Sesia, M., & Candès, E. J. (2020). A comparison of some conformal quantile regression methods. Stat, 9(1). https://doi.org/10.1002/sta4.261

Vancouver

Sesia M, Candès EJ. A comparison of some conformal quantile regression methods. Stat. 2020;9(1). doi:10.1002/sta4.261

BibTeX
@article{Sesia2020,
  author = {Sesia, Matteo and Candès, Emmanuel J.},
  title = {A comparison of some conformal quantile regression methods},
  journal = {Stat},
  year = {2020},
  volume = {9},
  number = {1},
  doi = {10.1002/sta4.261}
}
Start for free →