Practical Tools for Designing and Weighting Survey Samples

ISBN-13: 9783319936314 / Angielski / Twarda / 2018 / 776 str.

Richard Valliant; Jill A. Dever; Frauke Kreuter

Opis

This book explains approaches long used by survey statisticians, illustrating how existing software can be used to solve survey problems, and developing some specialized software where needed.

Recenzje

"This book attempts to explain long used approaches, illustrate how existing software can be used to solve survey problems, and develop some specialized software where needed with a focus on related practitioners (i.e. students, survey statisticians, and other survey practitioners). ... this book may be useful to students, survey statisticians, social scientists and other survey practitioners. The book may also serve as a useful reference for other professionals engaged in the conduct of sample surveys." (Sada Nand Dwivedi, ISCB News, iscb.info, Issue 67, June, 2019)

Spis treści

Preface

Acknowledgements

1 An Overview of Sample Design and Weighting

1.1 Background and Terminology

1.2 Chapter Guide

Part I Designing Single-Stage Sample Surveys

2 Project 1: Design a Single-Stage Personnel Survey

2.1 Specifications for the Study

2.2 Questions Posed by the Design Team

2.3 Preliminary Analyses

2.4 Documentation

2.5 Next Steps

3 Sample Design and Sample Size for Single-Stage Surveys 

3.1 Determining a Sample Size for a Single-Stage Design 

3.1.1 Simple Random Sampling

3.1.2 Stratified Simple Random Sampling

3.2 Finding Sample Sizes When Sampling with Varying Probabilities 

3.2.1 Probability Proportional to Size Sampling

3.2.2 Regression Estimates of Totals

3.3 Other Methods of Sampling

3.4 Estimating Population Parameters from a Sample

3.5 Special Topics

3.5.1 Rare Characteristics

3.5.2 Domain Estimates

3.6 More Discussion of Design Effects

3.7 Software for Sample Selection

3.7.1 R Packages

3.7.2 SAS PROC SURVEYSELECT

Exercises

4 Power Calculations and Sample Size Determination 

4.1 Terminology and One-Sample Tests

4.2 Power in a One-Sample Test

4.3 Two-Sample Tests

4.3.1 Differences in Means

4.3.2 Differences in Proportions

4.3.3 Special Case: Relative Risk

4.3.4 Special Case: Effect Sizes

4.4 R Power Functions

4.5 Power and Sample Size Calculations in SAS. 

Exercises

5 Mathematical Programming

5.1 Multicriteria Optimization

5.2 Microsoft Excel Solver

5.3 SAS PROC NLP

5.4 SAS PROC OPTMODEL

5.5 R Alabama Package

<

6 Outcome Rates and Effect on Sample Size

6.1 Disposition Codes

6.2 Definitions of Outcome Rates

6.3 Sample Units with Unknown AAPOR Classification

6.4 Weighted Versus Unweighted Rates

6.5 Accounting for Sample Losses in Determining Initial Sample Size

6.5.1 Sample Size Inflation Rates at Work

6.5.2 Replicates

Exercises

7 The Personnel Survey Design Project: One Solution 

7.1 Overview of the Project 

7.2 Formulate the Optimization Problem

7.2.1 Objective Function 

7.2.2 Decision Variables 

7.2.3 Optimization Parameters

7.2.4 Specified Survey Constraints 

7.3 One Solution 

7.3.1 Power Analyses

7.3.2 Optimization Results

7.4 Additional Sensitivity Analysis

7.5 Conclusion 

Part II Multistage Designs 

8 Project 2: Designing an Area Sample 

9 Designing Multistage Samples 

9.1 Types of PSUs 

9.2 Basic Variance Results 

9.2.1 Two-Stage Sampling 

9.2.2 Nonlinear Estimators in Two-Stage Sampling 

9.2.3 More General Two-Stage Designs 

9.2.4 Three-Stage Sampling 

9.3 Cost Functions and Optimal Allocations for Multistage Sampling 

9.3.1 Two-Stage Sampling When Numbers of Sample PSUs and Elements per PSU Are Adjustable 

9.3.2 Three-Stage Sampling When Sample Sizes Are Adjustable 

9.3.3 Two- and Three-Stage Sampling with a Fixed Set of PSUs 

9.4 Estimating Measures of Homogeneity and Variance Components

9.4.1 Two-Stage Sampling 

9.4.2 Three-Stage Sampling 

9.4.3 Using Anticipated Variances 

The lme4 R package has been updated so that the syntax in the 1st edition no longer works. We will revise the examples in this section for the new version of the package.

9.5 Stratification of PSUs 

9.6 Identifying Certainties 

Exercises

10 Area Sampling

10.1 Census Geographic Units

10.2 Census Data and American Community Survey Data

10.3 Units at Different Stages of Sampling

10.3.1 Primary Sampling Units

10.3.2 Secondary Sampling Units

10.3.3 Ultimate Sampling Units

10.4 Examples of Area Probability Samples

10.4.1 Current Population Survey

10.4.2 National Survey on Drug Use and Health

10.4.3 Panel Arbeitsmarkt und Soziale Sicherung

10.5 Composite MOS for Areas

10.5.1 Designing the Sample from Scratch

10.5.2 Using the Composite MOS with an Existing PSU Sample

10.6 Effects of Population Change: The New Construction Issue

10.7 Special Address Lists

10.7.1 Allocations in ABS using Mathematical Programming 

Mathematical programming allows efficient allocations to be made to domains (e.g., age groups) using information on housing units that can be purchased from commercial list makers.  Discussion and examples will be added to illustrate this technique.  The following article will be the basis for examples:

Valliant, R., Hubbard, F., Lee, S., Chang, W. (2014). “Efficient Use of Commercial Lists in Household Sampling”, Journal of Survey Statistics and Methodology, 2, 182-209.

Exercises

11 The Area Sample Design: One Solution

Part III Survey Weights and Analyses

12 Project 3: Weighting a Personnel Survey

13 Basic Steps in Weighting

13.1 Overview of Weighting

13.2 Theory of Weighting and Estimation

13.3 Base Weights

13.4 Adjustments for Unknown Eligibility

13.5 Adjustments for Nonresponse

13.5.1 Weighting Class Adjustments

13.5.2 Propensity Score Adjustments

13.5.3 Classification Algorithms

13.6 Collapsing Predefined Classes

13.7 Weighting for Multistage Designs

13.8 Next Steps in Weighting

Exercises

14 Calibration and Other Uses of Auxiliary Data in Weighting

14.1 Weight Calibration

14.2 Poststratified and Raking Estimators

14.3 GREG and Calibration Estimation

14.3.1 Links Between Models, Sample Designs, and Estimators-Special Cases

14.3.2 More General Examples

14.4 Weight Variability

14.4.1 Quantifying the Variability

14.4.2 Methods to Limit Variability

Exercises

15 Variance Estimation

15.1 Exact Methods

15.2 Linear Versus Nonlinear Estimators

15.3 Linearization Variance Estimation

15.3.1 Estimation Method

15.3.2 Confidence Intervals and Degrees of Freedom

15.3.3 Accounting for Non-negligible Sampling Fractions

15.3.4 Domain Estimation

15.3.5 Assumptions and Limitations

15.3.6 Special Cases: Poststratification and Quantiles

15.3.7 Handling Multiple Weighting Steps with Linearization

15.4 Replication

15.4.1 Jackknife Replication

15.4.2 Balanced Repeated Replication

15.4.3 Bootstrap

15.5 Combining PSUs or Strata

15.5.1 Combining to Reduce the Number of Replicates

15.5.2 How Many Groups and Which Strata and PSUs to Combine

15.5.3 Combining Strata in One-PSU-per-Stratum Designs

15.6 Handling Certainty PSUs

Exercises 

16 Weighting the Personnel Survey: One Solution

16.1 The Data Files

16.2 Base Weights

16.3 Disposition Codes and Mapping into Weighting Categories

16.4 Adjustment for Unknown Eligibility

16.5 Variables Available for Nonresponse Adjustment

16.6 Nonresponse Adjustments

16.7 Calibration to Population Counts

16.8 Writing Output Files

16.9 Example Tabulations

Part IV Other Topics

17 Multiphase Designs

17.1 What is a Multiphase Design?

17.2 Examples of Different Multiphase Designs

17.2.1 Double Sampling for Stratification

17.2.2 Nonrespondent Subsampling

17.2.3 Responsive Designs

17.2.4 General Multiphase Designs

17.3 Survey Weights

17.3.1 Base Weights

17.3.2 Analysis Weights

17.4 Estimation

17.4.1 Descriptive Point Estimation

17.4.2 Variance Estimation

17.4.3 Generalized Regression Estimator (GREG)

17.5 Design Choices

   17.5.1 Multiphase versus Single Phase

   17.5.2 Sample Size Calculations

17.6 R Software

Exercises

18. Non-probability Samples

18.1 Types of Non-probability Samples

18.2 Potential Problems

18.3 Quasi-randomization Approach

18.4 Superpopulation Modeling Approach

19 Process Control and Quality Measures

19.1 Design and Planning

19.2 Quality Control in Frame Creation and Sample Selection

19.3 Monitoring Data Collection . . .

19.4 Performance Rates and Indicators

19.5 Data Editing

19.5.1 Editing Disposition Codes

19.5.2 Editing the Weighting Variables

19.6 Quality Control of Weighting Steps

19.7 Specification Writing and Programming

19.8 Project Documentation and Archiving

 

Part V. Backmatter 

Appendix A: Notation Glossary

Appendix B: Data Sets

Appendix C: R Functions Used in this Book

1 r="" overview

C.2 Author-Defined R Functions

References

Solutions to Selected Exercises

Subject Index

O autorze

Richard Valliant, PhD, is Research Professor Emeritus at the Institute for Social Research at the University of Michigan and at the Joint Program in Survey Methodology at the University of Maryland. He is a Fellow of the American Statistical Association, an elected member of the International Statistical Institute, and has been an Associate Editor of the Journal of the American Statistical Association, Journal of Official Statistics, and Survey Methodology.   

Jill A. Dever, PhD, is Senior Research Statistician at RTI International in Washington, DC. She is a Fellow of the American Statistical Association, Associate Editor for Survey Methodology and the Journal of Official Statistics, and an Assistant Research Professor in the Joint Program in Survey Methodology at the University of Maryland. She has served on several panels for the National Academy of Sciences and as a task force member for the American Association of Public Opinion Research’s report on nonprobability sampling.

Frauke Kreuter, PhD, is Professor and Director of the Joint Program in Survey Methodology at the University of Maryland, Professor of Statistics and Methodology at the University of Mannheim, and Head of the Statistical Methods Research Department at the Institute for Employment Research (IAB) in Nürnberg, Germany. She is a Fellow of the American Statistical Association and has been Associate Editor of the Journal of the Royal Statistical Society, Journal of Official Statistics, Sociological Methods and Research, Survey Research Methods, Public Opinion Quarterly, American Sociological Review, and the Stata Journal. She is founder of the International Program for Survey and Data Science and co-founder of the Coleridge Initiative.

Blurb

The goal of this book is to put an array of tools at the fingertips of students, practitioners, and researchers by explaining approaches long used by survey statisticians, illustrating how existing software can be used to solve survey problems, and developing some specialized software where needed.  This volume serves at least three audiences: (1) students of applied sampling techniques; 2) practicing survey statisticians applying concepts learned in theoretical or applied sampling courses; and (3) social scientists and other survey practitioners who design, select, and weight survey samples.

The text thoroughly covers fundamental aspects of survey sampling, such as sample size calculation (with examples for both single- and multi-stage sample design) and weight computation, accompanied by software examples to facilitate implementation. Features include step-by-step instructions for calculating survey weights, extensive real-world examples and applications, and representative programming code in R, SAS, and other packages.

Since the publication of the first edition in 2013, there have been important developments in making inferences from nonprobability samples, in address-based sampling (ABS), and in the application of machine learning techniques for survey estimation. New to this revised and expanded edition:

•           Details on new functions in the PracTools package

•           Additional machine learning methods to form weighting classes

•           New coverage of nonlinear optimization algorithms for sample allocation

•           Reflecting effects of multiple weighting steps (nonresponse and calibration) on standard errors

•           A new chapter on nonprobability sampling

•           Additional examples, exercises, and updated references throughout

Richard Valliant, PhD, is Research Professor Emeritus at the Institute for Social Research at the University of Michigan and at the Joint Program in Survey Methodology at the University of Maryland. He is a Fellow of the American Statistical Association, an elected member of the International Statistical Institute, and has been an Associate Editor of the Journal of the American Statistical Association, Journal of Official Statistics, and Survey Methodology.   

Jill A. Dever, PhD, is Senior Research Statistician at RTI International in Washington, DC. She is a Fellow of the American Statistical Association, Associate Editor for Survey Methodology and the Journal of Official Statistics, and an Assistant Research Professor in the Joint Program in Survey Methodology at the University of Maryland. She has served on several panels for the National Academy of Sciences and as a task force member for the American Association of Public Opinion Research’s report on nonprobability sampling.

Frauke Kreuter, PhD, is Professor and Director of the Joint Program in Survey Methodology at the University of Maryland, Professor of Statistics and Methodology at the University of Mannheim, and Head of the Statistical Methods Research Department at the Institute for Employment Research (IAB) in Nürnberg, Germany. She is a Fellow of the American Statistical Association and has been Associate Editor of the Journal of the Royal Statistical Society,Journal of Official Statistics, Sociological Methods and Research, Survey Research Methods, Public Opinion Quarterly, American Sociological Review, and the Stata Journal. She is founder of the International Program for Survey and Data Science and co-founder of the Coleridge Initiative.