剑桥概率和统计：自助法及其应用

　　Wedecidedtotrytowriteabalanced account ofresamplingmethods，toincludebasic aspects of the theory which underpinned the methods，and to show as manyapplications as we could in order to illustrate the fun potential of the methods-warts and alL We quickly realized that in order for uS and others to understandand use the bootstrap，we would need suitable software，and producing it led usfurther towards a practically oriented treatment.0ur view was cemented by twofurther developments：the appearance oftwo excellent books，one bv PIeter Hallon the asymptotic theory and the other on basic methods bv Bradley Efron andRobert Tibshirani；and the chance to give further courses that included practicals.Our experience has been that hands-on computing is essential in coming to gripswith resampling ideas，so we have included practicals in this booL as well as moretheoretical problems.

Preface
1 Introduction
2 The Basic Bootstraps
2．1 Introduction
2．2 Parametric Simulation
2．3 Nonparametric Simulation
2．4 Simple Contidence Intervais
2．5 Redudng Error
2．6 Statistical Issues
2．7 Nonparametric Approximations for Variance and Bias
2．8 Subsampling Methods
2．9 Bibliographic Notes
2．10 Problems
2．11 Practicals

3 Further Idess
3．1 Introduction
3．2 Several Samples
3．3 Sereiparametric Models
3．4 Smooth Estimates of F
3．5 Censoring
3．6 Missing Data
3．7 Finite Population Sampling
3．8 Hierarchical Data
3．9 Bootstrapping the Bootstrap
3．10 Bootstrap Diagnostics
3．11 Choice of Estimator from the Data
3．12 Bibliographic Notes
3．13 Problems
3．14 Practicals

4 Tests
4．1 Introduction
4．2 Resampfing for Parametric TEsts
4．3 Nonparametric Permutafion Tests
4．4 Nonparametric Bootstrap Tests
4．5 Adjusted P-valnes
4．6 Estimating Properties of Tests
4．7 Bibliographic Notes
4．8 Problems
4．9 Practicals

5 Confidence latervals
5．1 Introduction
5．2 Basic Confidenee Limit Methods
5．3 Percentile Methods
5．4 Theorotical Comparison of Methods
5．5 Inversion of Significance Tcsts
5．6 Doublc Bootstrap Methods
5．7 Empirical Comparison of Boot~rap Method
5．8 Multiparameter Methods
5．9 Conditional Confidence Regions
5．10 Prediction
5．11 Bibliographic Notes
5．12 problems
5．13 Practicals

6 Liaear Regression
6．1 Introduction
6．2 Least Squares Linear Regression
6．3 Multiple Linear Regression
6．4 Aggregate Prediction Error and Variable Selection
6．5 Robust Regression
6．6 Bibliographic Notes
6．7 Problems
6．8 Pracdcals

7 Farther Topics in Regression
7．1 Introduction
7．2 Generalized Linear Modds
7．3 Survival Data
7．4 Other Nonlinear Models
7．5 Misclassifieation Error
7．6 Nonparametric Regression
7．7 Bibliographic Notes
7．8 Problems
7．9 Practicals

8 Complex Dependence
8．1 Introduction
8．2 Time Series
8．3 PointProcesses
8．4 Bibliographic Notes
8．5 Problems
8．6 Practicals

9 Improved Calculation
9．1 Introduction
9．2 Balanced Bootaraps
9．3 ControI Methods
9．4 Importance Resanling
9．5 Saddlepoint Approximation
9．6 Bibliographic Notes
9．7 Problems
9．8 Practicals

10 Semiparametrie Likelihood Inference
10．1 LiIcelihood
10．2 Multinoimal-Based Likelihoods
10．3 Bootstrap Likelihood
10．4 Likelihood Based on Confidence Sets
10．5 Bayesian Bootstraps
10．6 Bibfiographic Notes
10．7 Problems
10．8 Practicals

12 Computer Implementation
11．1 Introduction
11．2 Basic Bootstraps
11．3 Further Ideas
11．4 Tests
11．5 Confidence Intervals
11．6 Linear Regression
11．7 Further Topics in Regression
11．8 Time Series
11．9 Improved Simulation
11．10 Semiparametric Likelihoods

Appendix A．Cumulaat Calcalatioas
Bibliography
NameIndex
Exampleindex
Subject index

　　The publication in 1979 of BradIcy Efrons first article on bootstrap methods was amajor event in Statistics，at once synthesizing some of the cartier resampling ideasand establishing a new framework for simulation.based statistical analysis.The ideaof replacing complicated and often inaccurate approximations to biases，variances，and other measures of uncertainty by computer simulations caught the imaginationof both theoretical researchers and Users of statistical methods.Theoreticianssharpened their pencils and set about establishing mathematical conditions underwhich the idea could work.Once they had overcome their iuitial skepticism.appliedworkers sat down at their terminals and began to amass empirical evidence thatthe bootstrap often did work better than traditional methods.The early trickle ofPapers quickly became a torrent，with new additions to the literature appearingeverymonth nditwashardto seewhenwould be agoodmomenttotryto chartthe waters.Then the organizers of COMPSTAT 92 invited us to present a courseon the topic，and shortly afterwards we began to write this book.
　　Wedecidedtotrytowriteabalanced account ofresamplingmethods，toincludebasic aspects of the theory which underpinned the methods，and to show as manyapplications as we could in order to illustrate the fun potential of the methods-warts and alL We quickly realized that in order for uS and others to understandand use the bootstrap，we would need suitable software，and producing it led usfurther towards a practically oriented treatment.0ur view was cemented by twofurther developments：the appearance oftwo excellent books，one bv PIeter Hallon the asymptotic theory and the other on basic methods bv Bradley Efron andRobert Tibshirani；and the chance to give further courses that included practicals.Our experience has been that hands-on computing is essential in coming to gripswith resampling ideas，so we have included practicals in this booL as well as moretheoretical problems.
　　As the book fxpanded，we realized that a fully comprehensive treatment wasbeyond US，and that certain topics could be given only a cursory treatment becausetoo little is known about them.So it is that the reader will find only brief accountsof bootstrap methods for hierarchical data，missing data problems。model selection，robust estimation，nonparametric regression，and complex data.But we do try topoint the more ambitious reader in the fight direction.