Resampling Methods: A Practical Guide to Data Analysis, Third Edition
Author:
|
Phillip I. Good |
| Publisher: |
Birkhauser |
| Copyright: |
2006 |
| ISBN-13: |
978-0-8176-4386-7 |
| Pages: |
218; hardcover |
| Price: |
$56.50 |
|
|
|
|
Comment from the Stata technical group
Today’s more powerful computers have made data-resampling methods,
such as bootstrap, permutations, and cross-validation more popular. Although
computationally intensive, these methods of depicting sample-to-sample
variability are superior to their asymptotic counterparts because they lack
the stringent model assumptions that sometimes accompany the classically
used methods. This text is an interesting hybrid—part introductory
statistics text and part handbook on resampling methodology, with both areas
presented side by side. The classical introductory topics of point
estimation, testing, and classification are not presented in the usual way
but instead show how resampling methods may be used in their implementation.
More advanced topics, such as power and sample size, multivariate
statistics, experimental design, and model building are also covered, again
with an eye to how resampling methods may be incorporated into their use.
The material in this new edition has been reorganized to help readers get to
the material they are interested in quickly. Worked examples using Stata
are presented within the chapters. There is also a glossary of statistical
terms at the end of the book.
Table of contents
Preface to the Third Edition
Preface to the Second Edition
Preface to the First Edition
Which chapter should I read?
1 Software for resampling
1.1 Box Sampler©: An Excel Add-In
1.2 C++
1.3 CART®
1.4 EViews
1.5 MatLab®
1.6 R
1.7 Resampling Stats®
1.8 SAS®
1.9 S-PLUS®
1.10 Stata®
1.11 Statistical Calculator
1.12 StatXact®
1.13 XLminer®
1.14 Miscellaneous
2 Estimating population parameters
2.1 Population parameters
2.1.1 Frequency distribution and percentiles
2.1.2 Central values
2.1.3 Measures of dispersion
2.2 Samples and populations
2.2.1 The bootstrap
2.2.2 Programming the bootstrap
2.2.3 Estimating bias
2.2.4 An example
2.3 Confidence intervals
2.3.1 Limitations of the percentile bootstrap confidence interval
2.3.2 The bias-corrected bootstrap confidence interval
2.3.3 Computer code: the BCα bootstrap
2.3.4 The bootstrap-t
2.3.5 Parametric bootstrap
2.3.6 Smoothing the bootstrap
2.3.7 Importance sampling and the tilted bootstrap
2.3.8 Iterated bootstrap
2.4 Summary
2.5 To learn more
2.6 Exercises
3 Comparing two populations
3.1 A laboratory experiment
3.2 Analyzing the experiment
3.3 Some statistical considerations
3.3.1 Framing the hypothesis
3.3.2 Hypothesis versus alternative
3.3.3 Unpredictable variation
3.3.4 Some not-so-hidden assumptions
3.3.5 More general hypotheses
3.4 Computing the p-value
3.4.1 Monte Carlo
3.4.2 Program code
3.4.3 One-sided versus two-sided test
3.5 Matched pairs
3.6 Unequal variances
3.6.1 Underlying assumptions
3.7 Comparing variances
3.7.1 Unequal sample sizes
3.8 To learn more
3.9 Exercises
4 Choosing the best procedure
4.1 Why you need to read this chapter
4.2 Fundamental concepts
4.2.1 Two types of error
4.2.2 Losses
4.2.3 Significance level and power
4.2.4 What significance level should I use?
4.3 Confidence intervals
4.3.1 Interpreting the confidence interval
4.3.2 Multiple tests
4.4 Which test should be used?
4.4.1 Types of data
4.4.2 Assumptions
4.4.3 Recognizing common parametric distributions
4.4.4 Transformations
4.4.5 Distribution-free tests
4.4.6 Which test?
4.5 Summary
4.6 To learn more
4.7 Exercises
5 Experimental design and analysis
5.1 Separating signal from noise
5.1.1 Blocking
5.1.2 Analyzing a blocked experiment
5.1.3 Measuring factors we can’t control
5.1.4 Randomization
5.2 k-Sample Comparison
5.2.1 Testing for any and all differences
5.2.2 Analyzing a one-way table
5.2.3 Ordered alternatives
5.2.4 Calculating Pitman correlation
5.2.5 Effect of ties
5.3 Balanced designs
5.3.1 Main effects
5.3.2 Analyzing a two-way table
5.3.3 Testing for interactions
5.3.4 Synchronized rearrangements
5.3.5 A worked-through example
5.4 Designing an experiment
5.4.1 Latin square
5.5 Determining sample size
5.5.1 Group sequential designs
5.6 Unbalanced designs
5.6.1 Multidimensional contingency tables
5.6.2 Missing combinations
5.7 Summary
5.8 To learn more
5.9 Exercises
6 Categorical data
6.1 Fisher’s exact test
6.1.1 Computing Fisher’s exact test
6.1.2 One-tailed and two-tailed tests
6.1.3 The two-tailed test
6.1.4 When to accept
6.1.5 Is the sample large enough?
6.2 Odds ratio
6.2.1 Stratified 2 × 2’s
6.3 Exact significance levels
6.4 Unordered r × c contingency tables
6.4.1 Test of association
6.4.2 Causation versus association
6.5 Ordered statistical tables
6.5.1 More than two rows and two columns
6.6 Summary
6.7 To learn more
6.8 Exercises
7 Multiple variables and multiple hypotheses
7.1 Single-valued test statistic
7.1.1 Applications to repeated measures
7.1.2 An example
7.2 Combining univariate tests
7.3 The generalized quadratic sum
7.3.1 Mantel’s U
7.3.2 Example in epidemiology
7.3.3 Further generalization
7.3.4 The MRPP statistic
7.3.5 An example: blue grouse migration data
7.4 Multiple hypotheses
7.4.1 Testing for trend
7.5 Summary
7.6 To learn more
7.7 Exercises
8 Model building
8.1 Picturing relationships
8.2 Unpredictable variation
8.2.1 Building a model
8.2.2 Bivariate dependence
8.2.3 Confidence interval for the correlation coefficient
8.2.4 But does the model make sense?
8.2.5 Estimating the parameters
8.3 Linear regression
8.3.1 Other regression models
8.4 Improving the model
8.4.1 Testing for the significance in multipredictor regression
8.4.2 Comparing two regression lines
8.4.3 Prediction error
8.4.4 Correcting for bias
8.5 Validation
8.5.1 Metrics
8.5.2 Cross-validation
8.5.3 Using the bootstrap for model validation
8.6 Summary
8.7 To learn more
8.8 Exercises
9 Decision trees
9.1 Classification
9.2 Consumer survey
9.3 Trees versus regression
9.3.1 Head-to-head comparison
9.3.2 Which variables?
9.4 To learn more
9.5 Exercises
Answers to selected exercises
Bibliography
Glossary
Author index
Subject index
|
){kind=link}
){kind=link}