Multiple Imputation for Nonresponse in Surveys (Google eBook)

Front Cover
John Wiley & Sons, Sep 25, 2009 - Mathematics - 288 pages
0 Reviews
Demonstrates how nonresponse in sample surveys and censuses can be handled by replacing each missing value with two or more multiple imputations. Clearly illustrates the advantages of modern computing to such handle surveys, and demonstrates the benefit of this statistical technique for researchers who must analyze them. Also presents the background for Bayesian and frequentist theory. After establishing that only standard complete-data methods are needed to analyze a multiply-imputed set, the text evaluates procedures in general circumstances, outlining specific procedures for creating imputations in both the ignorable and nonignorable cases. Examples and exercises reinforce ideas, and the interplay of Bayesian and frequentist ideas presents a unified picture of modern statistics.
  

What people are saying - Write a review

We haven't found any reviews in the usual places.

Contents

1 INTRODUCTION
1
12 Examples of Surveys with Nonresponse
4
13 Properly Handling Nonresponse
7
14 Single Imputation
11
15 Multiple Imputation
15
16Numerical Example Using Multiple Imputation
19
17 Guidance for the Reader
22
Problems
23
42 General Conditions for the RandomizationValidity of Infinitem RepeatedImputation Inferences
116
43 Examples of Proper and Improper Imputation Methods in a Simple Case with Ignorable Nonresponse
120
44 Further Discussion of Proper Imputation Methods
125
45 The Asymptotic Distribution of QmUm Bo for Proper Imputation Methods
128
46 Evaluations of Finitem Inferences with Scalar Estimands
132
47 Evaluation of Significance Levels from the Moment Based Statistics Dm and Dm with Multicomponent Estimands
137
48 Evaluation of Significance Levels Based on Repeated Significance Levels
144
Problems
148

2 STATISTICAL BACKGROUND
27
22 Variables in the Finite Population
28
23 Probability Distributions and Related Calculations
31
24 Probability Specifications for Indicator Variables
35
25 Probability Specifications for X Y
39
26 Bayesian Inference for a Population Quantity
48
27 Interval Estimation
54
28 Bayesian Procedures for Constructing Interval Estimates Including Significance Levels and Point Estimates
59
29 Evaluating the Performance of Procedures
62
210 Similarity of Bayesian and RandomizationBased Inferences in Many Practical Cases
65
Problems
68
3 UNDERLYING BAYESIAN THEORY
75
32 Key Results for Analysis When the Multiple Imputations Are Repeated Draws from the Posterior Distribution of the Missing Values
81
33 Inference for Scalar Estimands from a Modest Number of Repeated CompletedData Means and Variances
87
34 Significance Levels for Multicomponent Estimands from a Modest Number of Repeated CompletedData Means and VarianceCovariance Matrices
94
35 Significance Levels from Repeated CompletedData Significance Levels
99
36 Relating the CompletedData and CompleteData Posterior Distributions When the Sampling Mechanism Is Ignorable
102
Problems
107
4 RANDOMIZATIONBASED EVALUATIONS
113
5 PROCEDURES WITH IGNORABLE NONRESPONSE
154
52 Creating Imputed Values under an Explicit Model
160
53 Some Explicit Imputation Models with Univariate yi and Covariates
166
54 Monotone Patterns of Missingness in Multivariate Yi
170
55 Missing Social Security Benefits in the Current Population Survey
178
56 Beyond Monotone Missingness
186
Problems
195
6 PROCEDURES WITH NONIGNORABLE NONRESPONSE
202
62 Nonignorable Nonresponse with Univariate yl and No X i
205
63 Formal Tasks with Nonignorable Nonresponse
210
64 Illustrating Mixture Modeling Using Educational Testing Service Data
215
65 Illustrating Selection Modeling Using CPS Data
222
66 Extensions to Surveys with FollowUps
229
67 FollowUp Response in a Survey of Drinking Behavior Among Men of Retirement Age
234
Problems
240
REFERENCES
244
AUTHOR INDEX
251
SUBJECT INDEX
253
Copyright

Common terms and phrases

About the author (2009)

Donald B. Rubin , PhD, is a John L. Loeb Professor of Statistics at Harvard University in Cambridge, MA.  He was named 2000-2001 Statistician of the Year by the Chicago Chapter of ASA.  His research interests include causal inference in experiments and observational studies, developing and applying statistical models to data in a variety of scientific disciplines, and the application of Bayesian and empirical Bayesian techniques.

Bibliographic information