Multimodel Inference

  title={Multimodel Inference},
  author={Kenneth P. Burnham and David R. Anderson},
  journal={Sociological Methods \& Research},
  pages={261 - 304}
The model selection literature has been generally poor at reflecting the deep foundations of the Akaike information criterion (AIC) and at making appropriate comparisons to the Bayesian information criterion (BIC). There is a clear philosophy, a sound criterion based in information theory, and a rigorous statistical foundation for AIC. AIC can be justified as Bayesian using a “savvy” prior on models that is a function of sample size and the number of model parameters. Furthermore, BIC can be… 

Model weights and the foundations of multimodel inference.

The usefulness of the weighted BIC (Bayesian information criterion) is suggested as a computationally simple alternative to AIC, based on explicit selection of prior model probabilities rather than acceptance of default priors associated with AIC.

Bayes Factors and Multimodel Inference

Noting the sensitivity of Bayes factors to the choice of priors on parameters, this work defines and proposes nonpreferential priors as offering a reasonable standard for objective multimodel inference.

Using Akaike’s information theoretic criterion in population

Akaike’s information-theoretic criterion for model discrimination (AIC) is often stated to “overfit”, i.e., it selects models with a higher dimension than the dimension of the model that generated

Normalized Information Criteria and Model Selection in the Presence of Missing Data

This work proposes a new approach that enables the use of classic well-known information criteria for model selection when there are missing data, and adjusts the current theory of information criteria through normalization, accounting for the different sample sizes used for each candidate model.

An evaluation of prior influence on the predictive ability of Bayesian model averaging

It is demonstrated that parsimonious priors may be favorable over priors that favor complexity for making predictions, and BMA performed better than a best single-model approach independently of the prior model weight for 6 out of 16 species.

A Conceptual Introduction to Bayesian Model Averaging

In this conceptual introduction, the principles of BMA are explained, its advantages over all-or-none model selection are described, and its utility is showcased in three examples: analysis of covariance, meta-analysis, and network analysis.

Multimodel Inference: Understanding AIC relative variable importance values

The goal of this material is to present extended theory and interpretation for the variable importance weights in multimodel information theoretic (IT) inference. We show that these statistics are a

Selecting Path Models in SEM: A Comparison of Model Selection Criteria

A comprehensive evaluation of various model selection criteria, including AIC, BIC, and their extensions, in selecting an optimal path model under a wide range of conditions over different compositions of candidate set, distinct values of misspecified parameters, and diverse sample sizes concludes that the BIC family in general outperformed AIC counterparts unless under small values of omitted parameters and sample sizes, where AIC performed better.

Bayesian Model Averaging as an Alternative to Model Selection for Multilevel Models

The Bayesian model averaging technique was investigated as an alternative method to the traditional model selection approaches for multilevel models (MLMs) through the Bayesian and the frequentist frameworks and suggested that BMA was a trustworthy alternative to traditional model comparison and selection approaches.



A Critique of the Bayesian Information Criterion for Model Selection

The Bayesian information criterion (BIC) has become a popular criterion for model selection in recent years. The BIC is intended to provide a measure of the weight of evidence favoring one model over

Model selection and multimodel inference : a practical information-theoretic approach

The second edition of this book is unique in that it focuses on methods for making formal statistical inference from all the models in an a priori set (Multi-Model Inference). A philosophy is

Model selection for extended quasi-likelihood models in small samples.

A small sample criterion (AICc) for the selection of extended quasi-likelihood models provides a more nearly unbiased estimator for the expected Kullback-Leibler information and often selects better models than AIC in small samples.

Bayesian Model Selection in Social Research

It is argued that P-values and the tests based upon them give unsatisfactory results, especially in large samples. It is shown that, in regression, when there are many candidate independent

Bayesian measures of model complexity and fit

The posterior mean deviance is suggested as a Bayesian measure of fit or adequacy, and the contributions of individual observations to the fit and complexity can give rise to a diagnostic plot of deviance residuals against leverages.

Generalizing the derivation of the schwarz information criterion

To better justify the widespread applicability of SIC, the criterion is derived in a very general framework: one which does not assume any specific form for the likelihood function, but only requires that it satisfies certain non-restrictive regularity conditions.

Approximate Bayes factors and accounting for model uncertainty in generalised linear models

A new approximation is proposed which uses only the output of standard computer programs for estimating generalised linear models, based on the Laplace method for integrals, and can be used to implement the Bayesian approach to accounting for model uncertainty.

Bayesian model averaging: a tutorial (with comments by M. Clyde, David Draper and E. I. George, and a rejoinder by the authors

Bayesian model averaging (BMA) provides a coherent mechanism for ac- counting for this model uncertainty and provides improved out-of- sample predictive performance.

Predictive Variable Selection in Generalized Linear Models

This prescription avoids the need to directly specify prior probabilities of models and prior densities for the parameters by proposing normal and conjugate priors for generalized linear models, each using a single prior prediction for the mean response to induce suitable priors to select a subset of variables.

Key Concepts in Model Selection: Performance and Generalizability.

  • E M Forster
  • Biology
    Journal of mathematical psychology
  • 2000
It seems that simplicity and parsimony may be an additional factor in managing these errors, in which case the standard methods of model selection are incomplete implementations of Occam's razor.