Galton, Edgeworth, Frisch, and prospects for quantile regression in econometrics

The work of three leading figures in the early history of econometrics is used to motivate some recent developments in the theory and application of quantile regression. We stress not only the robustness advantages of this form of semiparametric statistical method, but also the opportunity to recove...

Full description

Saved in:
Bibliographic Details
Published in:Journal of econometrics 2000-04, Vol.95 (2), p.347-374
Main Author: Koenker, Roger
Format: Article
Language:English
Subjects:
Applications
Applied sciences
Econometric models
Econometrics
Economic thought
Economists
Exact sciences and technology
Insurance, economics, finance
Interior point methods
Least absolute error regression
Linear inference, regression
Linear programming
Mathematical programming
Mathematics
Operational research and scientific management
Operational research. Management science
Probability and statistics
Quantile regression
Regression analysis
Regression depth
Sciences and techniques of general use
Statistics
Studies
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:The work of three leading figures in the early history of econometrics is used to motivate some recent developments in the theory and application of quantile regression. We stress not only the robustness advantages of this form of semiparametric statistical method, but also the opportunity to recover a more complete description of the statistical relationship between variables. A recent proposal for a more X-robust form of quantile regression based on maximal depth ideas is described along with an interesting historical antecedent. Finally, the notorious computational burden of median regression, and quantile regression more generally, is addressed. It is argued that recent developments in interior point methods for linear programming together with some new preprocessing ideas make it possible to compute quantile regressions as quickly as least-squares regressions throughout the entire range of problem sizes encountered in econometrics.
ISSN:0304-4076
1872-6895
DOI:10.1016/S0304-4076(99)00043-3