Carl-Erik Sarndal

Jan Wretman

Bengt Swensson

Model Assisted
Survey Sampling

Springer

Contents

Preface

PARTI
Principles of Estimation for Finite Populations and Important
Sampling Designs
CHAPTER 1
Survey Sampling in Theory and Practice
1.1
1.2
1.3
1.4
1.5
1.6
1.7
1.8
1.9
1.10

Surveys in Society
Skeleton Outline of a Survey
Probability Sampling
Sampling Frame
Area Frames and Similar Devices
Target Population and Frame Population
Survey Operations and Associated Sources of Error
Planning a Survey and the Need for Total Survey Design
Total Survey Design
The Role of Statistical Theory in Survey Sampling
Exercises

3
3
4
8
9
12
13
14
17
19
20
22

CHAPTER 2
Basic Ideas in Estimation from Probability Samples

24

2.1
2.2
2.3
2.4
2.5
2.6
2.7

24
24
27
30
33
36
38

Introduction
Population, Sample, and Sample Selection
Sampling Design
Inclusion Probabilities
The Notion of a Statistic
The Sample Membership Indicators
Estimators and Their Basic Statistical Properties

ix

x
2.8
2.9
2.10
2.11

Contents
The n Estimator and Its Properties
With-Replacement Sampling
The Design Effect
Confidence Intervals
Exercises

CHAPTER 3
Unbiased Estimation for Element Sampling Designs
3.1 Introduction
3.2 Bernoulli Sampling
3.3 Simple Random Sampling
3.3.1 Simple Random Sampling without Replacement
3.3.2 Simple Random Sampling with Replacement
3.4 Systematic Sampling
3.4.1 Definitions and Main Result
3.4.2 Controlling the Sample Size
3.4.3 The Efficiency of Systematic Sampling
3.4.4 Estimating the Variance
3.5 Poisson Sampling
3.6 Probability Proportional-to-Size Sampling
3.6.1 Introduction
3.6.2 nps Sampling
3.6.3 pps Sampling
3.6.4 Selection from Randomly Formed Groups
3.7 Stratified Sampling
3.7.1 Introduction
3.7.2 Notation, Definitions, and Estimation
3.7.3 Optimum Sample Allocation
3.7.4 Alternative Allocations under STSI Sampling
3.8 Sampling without Replacement versus Sampling with Replacement
3.8.1 Alternative Estimators for Simple Random Sampling with
Replacement
3.8.2 The Design Effect of Simple Random Sampling with Replacement
Exercises
CHAPTER 4
Unbiased Estimation for Cluster Sampling and Sampling in Two
' o r More Stages
4.1 Introduction
4.2 Single-Stage Cluster Sampling
4.2.1 Introduction
4.2.2 Simple Random Cluster Sampling
4.3 Two-Stage Sampling
4.3.1 Introduction
4.3.2 Two-Stage Element Sampling
4.4 Multistage Sampling
4.4.1 Introduction and a General Result
4.4.2 Three-Stage Element Sampling
4.5 With-Replacement Sampling of PSUs

42
48
53
55
58

61
61
62
66
66
72
73
73
76
78
83
85
87
87
90
97
99
100
100
101
104
106
110
110
112
114

124
124
126
126
129
133
133
135
144
144
146
150

Contents

xi

4.6 Comparing Simplified Variance Estimators in Multistage Sampling

Exercises

153
154

CHAPTER 5

Introduction to More Complex Estimation Problems

162

5.1
5.2
5.3
5.4
5.5
5.6
5.7
5.8
5.9
5.10

162
163
166
169
172
176
181
184
186
190
190
192
197
205
207

Introduction
The Effect of Bias on Confidence Statements
Consistency and Asymptotic Unbiasedness
n Estimators for Several Variables of Study
The Taylor Linearization Technique for Variance Estimation
Estimation of a Ratio
Estimation of a Population Mean
Estimation of a Domain Mean
Estimation of Variances and Covariances in a Finite Population
Estimation of Regression Coefficients
5.10.1 The Parameters of Interest
5.10.2 Estimation of the Regression Coefficients
5.11 Estimation of a Population Median
5.12 Demonstration of Result 5.10.1 ^
Exercises

PART II
Estimation through Linear Modeling, Using Auxiliary Variables
CHAPTER 6
The Regression Estimator
6.1
6.2
6.3
6.4
6.5
6.6
6.7
6.8

Introduction
Auxiliary Variables
The Difference Estimator
Introducing the Regression Estimator
Alternative Expressions for the Regression Estimator
The Variance of the Regression Estimator
Comments on the Role of the Model
Optimal Coefficients for the Difference Estimator
Exercises

219
219
219
221
225
230
234
238
239
242

CHAPTER 7
Regression Estimators for Element Sampling Designs

245

7.1 Introduction
7.2 Preliminary Considerations
7.3 The Common Ratio Model and the Ratio Estimator
7.3.1 The Ratio Estimator under SI Sampling
7.3.2 The Ratio Estimator under Other Designs
7.3.3 Optimal Sampling Design for the n Weighted Ratio Estimator
7.3.4 Alternative Ratio Models
7.4 The Common Mean Model
7.5 Models Involving Population Groups
7.6 The Group Mean Model and the Poststratified Estimator
7.7 The Group Ratio Model and the Separate Ratio Estimator

245
245
247
249
252
253
255
258
260
264
269

Contents
7.8
7.9

7.10

7.11
7.12
7.13

Simple Regression Models and Simple Regression Estimators

Estimators Based on Multiple Regression Models
7.9.1 Multiple Regression Models
7.9.2 Analysis of Variance Models
Conditional Confidence Intervals
7.10.1 Conditional Analysis for BE Sampling
7.10.2 Conditional Analysis for the Poststratification Estimator
Regression Estimators for Variable-Size Sampling Designs
A Class of Regression Estimators
Regression Estimation of a Ratio of Population Totals
Exercises

272
275
276
281
283
284
287
289
291
294
297

CHAPTER 8

Regression Estimators for Cluster Sampling and Two-Stage Sampling 303

8.1
8.2
8.3
8.4
8.5
8.6
8.7
8.8
8.9
8.10
8.11
8.12

Introduction
^
The Nature of the Auxiliary Information When Clusters of Elements
Are Selected
Comments on Variance and Variance Estimation in Two-Stage
Sampling
Regression Estimators Arising Out of Modeling at the Cluster Level
The Common Ratio Model for Cluster Totals
Estimation of the Population Mean When Clusters Are Sampled
Design Effects for Single-Stage Cluster Sampling
Stratified Clusters and Poststratified Clusters
Regression Estimators Arising Out of Modeling at the Element Level
Ratio Models for Elements
The Group Ratio Model for Elements
The Ratio Model Applied within a Single PSU
Exercises

303
304
307
308
312
314
315
319
322
327
330
332
333

PART III
Further Questions in Design and Analysis of Surveys
CHAPTER 9
Two-Phase Sampling
9.1
' 9.2
9.3
9.4
9.5
9.6
9.7
9.8
9.9

Introduction
Notation and Choice of Estimator
The 7i* Estimator
,
Two-Phase Sampling for Stratification
Auxiliary Variables for Selection in Two Phases
Difference Estimators
Regression Estimators for Two-Phase Sampling
Stratified Bernoulli Sampling in Phase Two
Sampling on Two Occasions
9.9.1 Estimating the Current Total
9.9.2 Estimating the Previous Total
9.9.3 Estimating the Absolute Change and the Sum of the Totals
Exercises

343
343
345
347
350
354
356
359
366
368
370
376
377
379

Contents

xiii

CHAPTER 10
Estimation for D o m a i n s

386

10.1
10.2
10.3
10.4
10.5
10.6
10.7
10.8
10.9

Introduction
The Background for Domain Estimation
The Basic Estimation Methods for Domains
Conditioning on the Domain Sample Size
Regression Estimators for Domains
A Ratio Model for Each Domain
Group Models for Domains
Problems Arising for Small Domains; Synthetic Estimation
More on the Comparison of Two Domains
Exercises

CHAPTER 11
Variance Estimation

386
387
390
396
397
403
405
408
412
413

418

11.1 Introduction
11.2 A Simplified Variance Estimator under Sampling without Replacement
11.3 The Random Groups Technique
11.3.1 Independent Random Groups
11.3.2 Dependent Random Groups
11.4 Balanced Half-Samples
11.5 The Jackknife Technique
11.6 The Bootstrap
11.7 Concluding Remarks
Exercises

418
421
423
423
426
430
437
442
444
445

CHAPTER 12
Searching for Optimal Sampling Designs

447

12.1
12.2
12.3
12.4
12.5
12.6
12.7
12.8

Introduction
Model-Based Optimal Design for the General Regression Estimator
Model-Based Optimal Design for the Group Mean Model
Model-Based Stratified Sampling
Applications of Model-Based Stratification
Other Approaches to Efficient Stratification
Allocation Problems in Stratified Random Sampling
Allocation Problems in Two-Stage Sampling
12.8.1 The n Estimator of the Population Total
12.8.2 Estimation of the Population Mean
12.9 Allocation in Two-Phase Sampling for Stratification
12.10 A Further Comment on Mathematical Programming
12.11 Sampling Design and Experimental Design
Exercises

447
448
455
456
461
462
465
471
471
475
478
480
481
481

CHAPTER 13
Further Statistical Techniques for Survey D a t a

485

13.1 Introduction
13.2 Finite Population Parameters in Multivariate Regression and
Correlation Analysis

485
486

xiv

Contents

13.3 The Effect of Sampling Design on a Statistical Analysis

13.4 Variances and Estimated Variances for Complex Analyses
13.5 Analysis of Categorical Data for Finite Populations
13.5.1 Test of Homogeneity for Two Populations
13.5.2 Testing Homogeneity for More than Two Finite Populations
13.5.3 Discussion of Categorical Data Tests for Finite Populations
13.6 Types of Inference When a Finite Population Is Sampled
Exercises

491
494
500
500
507
510
513
520

PART IV

A Broader View of Errors in Surveys

CHAPTER 14
Nonsampling Errors and Extensions of Probability Sampling Theory

525

14.1 Introduction
14.2 Historic Notes: The Evolution of the Probability Sampling Approach
14.3 Measurable Sampling Designs
14.4 Some Nonprobability Sampling Methods
14.5 Model-Based Inference from Survey Samples
14.6 Imperfections in the Survey Operations
14.6.1 Ideal Conditions for the Probability Sampling Approach
14.6.2 Extension of the Probability Sampling Approach
14.7 Sampling Frames
14.7.1 Frame Imperfections
14.7.2 Estimation in the Presence of Frame Imperfections
14.7.3 Multiple Frames
14.7.4 Frame Construction and Maintenance
14.8 Measurement and Data Collection
14.9 Data Processing
14.10 Nonresponse
Exercises

525
525
527
529
533
537
537
538
540
540
543
545
545
546
548
551
553

CHAPTER 15
Nonresponse

556

15.1 Introduction
15.2 Characteristics of Nonresponse
15.2.1 Definition of Nonresponse
15.2.2 Response Sets
15.2.3 Lack of Unbiased Estimators
15.3 Measuring Nonresponse
15.4 Dealing with Nonresponse
15.4.1 Planning of the Survey
15.4.2 Callbacks and Follow-Ups
15.4.3 Subsampling of Nonrespondents
15.4.4 Randomized Response
15.5 Perspectives on Nonresponse
15.6 Estimation in the Presence of Unit Nonresponse
15.6.1 Response Modeling

556
556
556
557
558
559
563
564
564
566
570
573
575
575

Contents

xv

15.6.2 A Useful Response Model

15.6.3 Estimators That Use Weighting Only
15.6.4 Estimators That Use Weighting as Well as Auxiliary Variables
15.7 Imputation
Exercises

577
580
583
589
595

CHAPTER 16
Measurement Errors

601

16.1
16.2
16.3
16.4
16.5
16.6
16.7
16.8
16.9
16.10
16.11

Introduction
On the Nature of Measurement Errors
The Simple Measurement Model
Decomposition of the Mean Square Error
.
^
The Risk of Underestimating the Total Variance
Repeated Measurements as a Tool in Variance Estimation
Measurement Models Taking Interviewer Effects into Account
Deterministic Assignment of Interviewers
Random Assignment of Interviewers to Groups
Interpenetrating Subsamples
A Measurement Model with Sample-Dependent Moments
Exercises

CHAPTER 17
Quality Declarations for Survey D a t a
17.1 Introduction
17.2 Policies Concerning Information on Data Quality
17.3 Statistics Canada's Policy on Informing Users of Data Quality and
Methodology
Exercise

601
602
605
608
612
614
617
618
622
627
630
634

637
637
638
641
648

APPENDIX A
Principles of N o t a t i o n

649

APPENDIX B
The MU284 Population

652

APPENDIX C
The Clustered M U 2 8 4 Population

660

APPENDIX D

The CO 124 Population

662

References
Answers to Selected Exercises
Author Index

666
680
684

Subject Index

688