Asymptotic bootstrap corrections of AIC for linear regression models

Authors:
Abd-Krim Seghouane
Affiliations:
National ICT Australia, Canberra Research Laboratory and The Australian National University, College of Engineering and Computer Science, Canberra, Australia
Venue:
Signal Processing
Year:
2010

Citing 9
Cited 2

Statistical aspects of model selection

From data to model
Segmenting Images Corrupted by Correlated Noise

IEEE Transactions on Pattern Analysis and Machine Intelligence
Fast communication: Bayesian estimation of the number of principal components

Signal Processing
Asymptotic MAP criteria for model selection

IEEE Transactions on Signal Processing
A small sample model selection criterion based on Kullback's symmetric divergence

IEEE Transactions on Signal Processing
Blind channel approximation: effective channel order determination

IEEE Transactions on Signal Processing
Paper: Modeling by shortest data description

Automatica (Journal of IFAC)
Minimax description length for signal denoising and optimized representation

IEEE Transactions on Information Theory
The AIC Criterion and Symmetrizing the Kullback–Leibler Divergence

IEEE Transactions on Neural Networks

Review: Variable selection in linear regression: Several approaches based on normalized maximum likelihood

Signal Processing
Identification of directed influence: Granger causality, kullback-leibler divergence, and complexity

Neural Computation

Quantified Score

Hi-index	0.08

Visualization

Abstract

The Akaike information criterion, AIC, and its corrected version, AIC"c are two methods for selecting normal linear regression models. Both criteria were designed as estimators of the expected Kullback-Leibler information between the model generating the data and the approximating candidate model. In this paper, two new corrected variants of AIC are derived for the purpose of small sample linear regression model selection. The proposed variants of AIC are based on asymptotic approximation of bootstrap type estimates of Kullback-Leibler information. These new variants are of particular interest when the use of bootstrap is not really justified in terms of the required calculations. As its the case for AIC"c, these new variants are asymptotically equivalent to AIC. Simulation results which illustrate better performance of the proposed AIC corrections when applied to polynomial regression in comparison to AIC, AIC"c and other criteria are presented. Asymptotic justifications for the proposed criteria are provided in the Appendix.