Bias-variance trade-off for prequential model list selection

Ernest Fokoue, Bertrand Clarke

Research output: Contribution to journalArticle

3 Citations (Scopus)

Abstract

The prequential approach to statistics leads naturally to model list selection because the sequential reformulation of the problem is a guided search over model lists drawn from a model space. That is, continually updating the action space of a decision problem to achieve optimal prediction forces the collection of models under consideration to grow neither too fast nor too slow to avoid excess variance and excess bias, respectively. At the same time, the goal of good predictive performance forces the search over good predictors formed from a model list to close in on the data generator. Taken together, prequential model list re-selection favors model lists which provide an effective approximation to the data generator but do so by making the approximation match the unknown function on important regions as determined by empirical bias and variance.

Original languageEnglish (US)
Pages (from-to)813-833
Number of pages21
JournalStatistical Papers
Volume52
Issue number4
DOIs
StatePublished - Nov 1 2011

Fingerprint

Trade-offs
Excess
Model
Generator
Optimal Prediction
Selection Model
Approximation
Reformulation
Decision problem
Updating
Predictors
Statistics
Unknown

Keywords

  • Bayes model averaging
  • Bias-variance trade-off
  • Model list selection
  • Model selection
  • Online prediction
  • Prequential

ASJC Scopus subject areas

  • Statistics and Probability
  • Statistics, Probability and Uncertainty

Cite this

Bias-variance trade-off for prequential model list selection. / Fokoue, Ernest; Clarke, Bertrand.

In: Statistical Papers, Vol. 52, No. 4, 01.11.2011, p. 813-833.

Research output: Contribution to journalArticle

Fokoue, Ernest ; Clarke, Bertrand. / Bias-variance trade-off for prequential model list selection. In: Statistical Papers. 2011 ; Vol. 52, No. 4. pp. 813-833.
@article{0bb51d6b986d4c11b33294ebb1f54bb6,
title = "Bias-variance trade-off for prequential model list selection",
abstract = "The prequential approach to statistics leads naturally to model list selection because the sequential reformulation of the problem is a guided search over model lists drawn from a model space. That is, continually updating the action space of a decision problem to achieve optimal prediction forces the collection of models under consideration to grow neither too fast nor too slow to avoid excess variance and excess bias, respectively. At the same time, the goal of good predictive performance forces the search over good predictors formed from a model list to close in on the data generator. Taken together, prequential model list re-selection favors model lists which provide an effective approximation to the data generator but do so by making the approximation match the unknown function on important regions as determined by empirical bias and variance.",
keywords = "Bayes model averaging, Bias-variance trade-off, Model list selection, Model selection, Online prediction, Prequential",
author = "Ernest Fokoue and Bertrand Clarke",
year = "2011",
month = "11",
day = "1",
doi = "10.1007/s00362-009-0289-6",
language = "English (US)",
volume = "52",
pages = "813--833",
journal = "Statistical Papers",
issn = "0932-5026",
publisher = "Springer New York",
number = "4",

}

TY - JOUR

T1 - Bias-variance trade-off for prequential model list selection

AU - Fokoue, Ernest

AU - Clarke, Bertrand

PY - 2011/11/1

Y1 - 2011/11/1

N2 - The prequential approach to statistics leads naturally to model list selection because the sequential reformulation of the problem is a guided search over model lists drawn from a model space. That is, continually updating the action space of a decision problem to achieve optimal prediction forces the collection of models under consideration to grow neither too fast nor too slow to avoid excess variance and excess bias, respectively. At the same time, the goal of good predictive performance forces the search over good predictors formed from a model list to close in on the data generator. Taken together, prequential model list re-selection favors model lists which provide an effective approximation to the data generator but do so by making the approximation match the unknown function on important regions as determined by empirical bias and variance.

AB - The prequential approach to statistics leads naturally to model list selection because the sequential reformulation of the problem is a guided search over model lists drawn from a model space. That is, continually updating the action space of a decision problem to achieve optimal prediction forces the collection of models under consideration to grow neither too fast nor too slow to avoid excess variance and excess bias, respectively. At the same time, the goal of good predictive performance forces the search over good predictors formed from a model list to close in on the data generator. Taken together, prequential model list re-selection favors model lists which provide an effective approximation to the data generator but do so by making the approximation match the unknown function on important regions as determined by empirical bias and variance.

KW - Bayes model averaging

KW - Bias-variance trade-off

KW - Model list selection

KW - Model selection

KW - Online prediction

KW - Prequential

UR - http://www.scopus.com/inward/record.url?scp=80053904124&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=80053904124&partnerID=8YFLogxK

U2 - 10.1007/s00362-009-0289-6

DO - 10.1007/s00362-009-0289-6

M3 - Article

AN - SCOPUS:80053904124

VL - 52

SP - 813

EP - 833

JO - Statistical Papers

JF - Statistical Papers

SN - 0932-5026

IS - 4

ER -