Statistical Science
- Statist. Sci.
- Volume 34, Number 4 (2019), 545-565.
Models as Approximations II: A Model-Free Theory of Parametric Regression
Andreas Buja, Lawrence Brown, Arun Kumar Kuchibhotla, Richard Berk, Edward George, and Linda Zhao
Abstract
We develop a model-free theory of general types of parametric regression for i.i.d. observations. The theory replaces the parameters of parametric models with statistical functionals, to be called “regression functionals,” defined on large nonparametric classes of joint ${x\textrm{-}y}$ distributions, without assuming a correct model. Parametric models are reduced to heuristics to suggest plausible objective functions. An example of a regression functional is the vector of slopes of linear equations fitted by OLS to largely arbitrary ${x\textrm{-}y}$ distributions, without assuming a linear model (see Part I). More generally, regression functionals can be defined by minimizing objective functions, solving estimating equations, or with ad hoc constructions. In this framework, it is possible to achieve the following: (1) define a notion of “well-specification” for regression functionals that replaces the notion of correct specification of models, (2) propose a well-specification diagnostic for regression functionals based on reweighting distributions and data, (3) decompose sampling variability of regression functionals into two sources, one due to the conditional response distribution and another due to the regressor distribution interacting with misspecification, both of order $N^{-1/2}$, (4) exhibit plug-in/sandwich estimators of standard error as limit cases of ${x\textrm{-}y}$ bootstrap estimators, and (5) provide theoretical heuristics to indicate that ${x\textrm{-}y}$ bootstrap standard errors may generally be preferred over sandwich estimators.
Article information
Source
Statist. Sci., Volume 34, Number 4 (2019), 545-565.
Dates
First available in Project Euclid: 8 January 2020
Permanent link to this document
https://projecteuclid.org/euclid.ss/1578474017
Digital Object Identifier
doi:10.1214/18-STS694
Mathematical Reviews number (MathSciNet)
MR4048583
Keywords
Ancillarity of regressors misspecification econometrics sandwich estimator bootstrap bagging
Citation
Buja, Andreas; Brown, Lawrence; Kuchibhotla, Arun Kumar; Berk, Richard; George, Edward; Zhao, Linda. Models as Approximations II: A Model-Free Theory of Parametric Regression. Statist. Sci. 34 (2019), no. 4, 545--565. doi:10.1214/18-STS694. https://projecteuclid.org/euclid.ss/1578474017
See also
- Discussion of Models as Approximations I & II. Digital Object Identifier: doi:10.1214/19-STS722
- Comment on Models as Approximations, Parts I and II, by Buja et al. Digital Object Identifier: doi:10.1214/19-STS723
- Comment: Models as Approximations. Digital Object Identifier: doi:10.1214/19-STS724
- Discussion of Models as Approximations I & II. Digital Object Identifier: doi:10.1214/19-STS725
- Comment: "Models as Approximations I: Consequences Illustrated with Linear Regression" by A. Buja, R. Berk, L. Brown, E. George, E. Pitkin, L. Zhan and K. Zhang. Digital Object Identifier: doi:10.1214/19-STS726
- Comment: Models Are Approximations!. Digital Object Identifier: doi:10.1214/19-STS746
- Comment: Models as (Deliberate) Approximations. Digital Object Identifier: doi:10.1214/19-STS747
- Comment: Statistical Inference from a Predictive Perspective. Digital Object Identifier: doi:10.1214/19-STS748
- Discussion: Models as Approximations. Digital Object Identifier: doi:10.1214/19-STS756
- Models as Approximations—Rejoinder. Digital Object Identifier: doi:10.1214/19-STS762
Supplemental materials
- Supplement to “Models as Approximations II: A Model-Free Theory of Parametric Regression”. This supplement contains Appendices A-G.Digital Object Identifier: doi:10.1214/18-STS694SUPPSupplemental files are immediately available to subscribers. Non-subscribers gain access to supplemental files with the purchase of the article.
