Godfrey Leslie. Bootstrap Tests for Regression Models

London: Palgrave McMiillan, 2009. — 343 p.This volume contains an accessible discussion examining computationally-intensive techniques and bootstrap methods, providing ways to improve the finite-sample performance of well-known asymptotic tests for regression models. The book uses the linear regression model as a framework for introducing simulation-based tests to help perform econometric analyses.Tests for the classical linear regression model
Tests for linear regression models under weaker assumptions: random regressors and non-Normal IID errors
Tests for generalized linear regression models
HCCME-based tests
HAC-based tests
Finite-sample properties of asymptotic tests
Testing the significance of a subset of regressors
Testing for non-Normality of the errors
Using heteroskedasticity-robust tests of significance
Non-standard tests for linear regression models
Summary and concluding remarksSome key concepts and simple examples of tests for IID variables
Monte Carlo tests
Bootstrap tests
The classical Normal model
Models with IID errors from an unspecified distribution
Dynamic regression models and bootstrap schemes
The choice of the number of artificial samples
Asymptotic properties of bootstrap tests
The double bootstrap
Summary and concluding remarksA Monte Carlo test of the assumption of Normality
Simulation-based tests for heteroskedasticity
Monte Carlo tests for heteroskedasticity
Bootstrap tests for heteroskedasticity
Simulation experiments and tests for heteroskedasticity
Regression models with strictly exogenous regressors
Stable dynamic regression models
Some simulation evidence concerning asymptotic and bootstrap F tests
Bootstrapping LM tests for serial correlation in dynamic regression models
Restricted or unrestricted estimates as parameters of bootstrap worlds
Some simulation evidence on the choice between restricted and unrestricted estimates
Summary and concluding remarksAsymptotic analysis for predictive test statistics
Single and double bootstraps for predictive tests
Simulation experiments and results
Dynamic regression models
Using bootstrap methods with a battery of OLS diagnostic tests
Regression models and diagnostic tests
Bootstrapping the minimum p-value of several diagnostic test statistics
Simulation experiments and results
Bootstrapping tests for structural breaks
Testing constant coefficients against an alternative with an unknown breakpoint
Simulation evidence for asymptotic and bootstrap tests
Summary and conclusionsBootstrap methods for independent heteroskedastic errors
Model-based bootstraps
Pairs bootstraps
Wild bootstraps
Estimating function bootstraps
Bootstrapping dynamic regression models
Bootstrap methods for homoskedastic autocorrelated errors
Model-based bootstraps
Block bootstraps
Sieve bootstraps
Other methods
Asymptotic theory tests
Block bootstraps
Other methods
Summary and concluding remarksThe forms of test statistics
Simulation experiments
Bootstrapping heteroskedasticity-robust autocorrelation tests for dynamic models
The forms of test statistics
Simulation experiments
Bootstrapping heteroskedasticity-robust structural break tests with an unknown breakpoint
The forms of test statistics
Simulation experiments
Summary and conclusionsAsymptotic tests for models with non-nested regressors
Cox-type LLR tests
Artificial regression tests
Regularity conditions and orthogonal regressors
Testing with multiple alternatives
Tests for model selection
Evidence from simulation experiments
One non-nested alternative regression model: significance levels
One non-nested alternative regression model: power
One non-nested alternative regression model: extreme cases
Two non-nested alternative regression models: significance levels
Two non-nested alternative regression models: power
Bootstrapping the LLR statistic with non-nested models
Summary and concluding remarks
Epilogue