Jump to ContentJump to Main Navigation
Empirical Model Discovery and Theory EvaluationAutomatic Selection Methods in Econometrics$
Users without a subscription are not able to see the full content.

David F. Hendry and Jurgen A. Doornik

Print publication date: 2014

Print ISBN-13: 9780262028356

Published to MIT Press Scholarship Online: January 2015

DOI: 10.7551/mitpress/9780262028356.001.0001

Show Summary Details
Page of

PRINTED FROM MIT PRESS SCHOLARSHIP ONLINE (www.mitpress.universitypressscholarship.com). (c) Copyright The MIT Press, 2022. All Rights Reserved. An individual user may print out a PDF of a single chapter of a monograph in MITSO for personal use.date: 28 June 2022

Background to Automatic Model Selection

Background to Automatic Model Selection

(p.31) 3 Background to Automatic Model Selection
Empirical Model Discovery and Theory Evaluation

David F. Hendry

The MIT Press

This chapter explains the background in more detail, to provide a framework for the analysis in the rest of the book. We first note the many criticisms of model selection that have been made. While many of these are valid for some approaches, we will show that almost all are rebutted for general-to-specific (Gets) model selection. Many criticisms are based on an assumed level of knowledge where discovery is unnecessary, so fail to address that key issue. The main aim of this chapter is towalk through the six stages leading from simple selection to model discovery in a context where there are more candidate variables than observations. Part II will discuss these stages in detail, after chapter 5 considers the choice of evaluation criteria for selection methods, chapter 6 outlines the theory of reduction, and chapter 7 describes Gets. Here, we commence by considering a baseline Gets approach, denoted 1-cut, which can select a model from any number of candidate variables with just one decision for mutually orthogonal, valid conditioning variables in a constant-parameter setting given a sufficiently large sample. Next, we show how to obtain near unbiased estimates after selection, then compare the 1-cut method with the multi-path search approach embodied in Autometrics, gradually extending the analysis to include diagnostic checking for the selection being well-specified, then parsimonious-encompassing tests against the initial general formulation, as well as efficiently handling more candidate variables than observations in the special case where an indicator is allowed for every data point. We also note the specific issues of selecting lag lengths, handling integrated data, collinearity, evaluating the reliability of the finally selected model by simulation, data accuracy, and the implementation of theory. We conclude this chapter by sketching the extensions that are the focus of part III, including handling more variables than observations in general, automatically testing for multiple breaks and for super exogeneity, as well as modeling non-linear equations, ending with selecting forecasting models.

Keywords:   1-cut selection, multi-path search, parsimonious encompassing, selection bias

MIT Press Scholarship Online requires a subscription or purchase to access the full text of books within the service. Public users can however freely search the site and view the abstracts and keywords for each book and chapter.

Please, subscribe or login to access full text content.

If you think you should have access to this title, please contact your librarian.

To troubleshoot, please check our FAQs, and if you can't find the answer there, please contact us.