Complex survey data analysis with SAS

**Complex Survey Data Analysis with SAS®** is an invaluable resource for applied researchers analyzing data generated from a sample design involving any combination of stratification, clustering, unequal weights, or finite population correction factors. After clearly explaining how the presence of these features can invalidate the assumptions underlying most traditional statistical techniques, this book equips readers with the knowledge to confidently account for them during the estimation and inference process by employing the SURVEY family of SAS/STAT® procedures. The book offers comprehensive coverage of the most essential topics, including: * Drawing random samples * Descriptive statistics for continuous and categorical variables * Fitting and interpreting linear and logistic regression models * Survival analysis * Domain estimation * Replication variance estimation methods * Weight adjustment and imputation methods for handling missing data The easy-to-follow examples are drawn from real-world survey data sets spanning multiple disciplines, all of which can be downloaded for free along with syntax files from the author’s website: http://mason.gmu.edu/~tlewis18/. While other books may touch on some of the same issues and nuances of complex survey data analysis, none features SAS exclusively and as exhaustively. Another unique aspect of this book is its abundance of handy workarounds for certain techniques not yet supported as of SAS Version 9.4, such as the ratio estimator for a total and the bootstrap for variance estimation. **Taylor H. Lewis** is a PhD graduate of the Joint Program in Survey Methodology at the University of Maryland, College Park, and an adjunct professor in the George Mason University Department of Statistics. An avid SAS user for 15 years, he is a SAS Certified Advanced programmer and a nationally recognized SAS educator who has produced dozens of papers and workshops illustrating how to efficiently and effectively conduct statistical analyses using SAS. Cover 1 Half Title 2 Title Page 4 Copyright Page 5 Dedication 6 Table of Contents 8 Preface 12 Author 14 1: Features and Examples of Complex Surveys 16 1:1 Introduction 16 1:2 Definitions and Terminology 18 1:3 Overview of SAS/STAT Procedures Available to Analyze Survey Data 23 1:4 Four Features of Complex Surveys 24 1:4:1 A Hypothetical Expenditure Survey 24 1:4:2 Finite Population Corrections 25 1:4:3 Stratification 28 1:4:4 Clustering 31 1:4:5 Unequal Weights 38 1:4:6 Brief Summary of the Features of Complex Survey Data 41 1:5 Examples of Complex Surveys 42 1:5:1 Introduction 42 1:5:2 The National Ambulatory Medical Care Survey 43 1:5:3 The National Survey of Family Growth 43 1:5:4 Commercial Buildings Energy Consumption Survey 44 1:6 Summary 45 2: Drawing Random Samples Using PROC SURVEYSELECT 48 2:1 Introduction 48 2:2 Fundamental Sampling Techniques 49 2:2:1 Simple Random Sampling 49 2:2:2 Systematic Sampling 50 2:2:3 Probability Proportional to Size Sampling 52 2:3 Stratified Sampling 54 2:4 Cluster Sampling 57 2:5 Summary 60 3: Analyzing Continuous Variables Using PROC SURVEYMEANS 62 3:1 Introduction 62 3:2 Totals 62 3:3 Means 66 3:4 Ratios 70 3:5 Quantiles 74 3:6 Summary 78 4: Analyzing Categorical Variables Using PROC SURVEYFREQ 80 4:1 Introduction 80 4:2 Univariate Analyses 81 4:2:1 Descriptive Statistics 81 4:2:2 Alternative Methods of Constructing Confidence Intervals for Extreme Proportions 85 4:2:3 Goodness-of-Fit Tests 87 4:3 Bivariate Analyses 91 4:3:1 Introduction 91 4:3:2 Tests of Association 93 4:3:3 Risk Statistics and Odds Ratios 96 4:4 Multiway Tables 100 4:5 Summary 104 5: Fitting Linear Regression Models Using PROC SURVEYREG 106 5:1 Introduction 106 5:2 Linear Regression in a Simple Random Sampling Setting 107 5:3 Linear Regression with Complex Survey Data 117 5:4 Testing for a Reduced Model 125 5:5 Computing Unit-Level Statistics 126 5:6 Summary 132 6: Fitting Logistic Regression Models Using PROC SURVEYLOGISTIC 134 6:1 Introduction 134 6:2 Logistic Regression in a Simple Random Sampling Setting 135 6:3 Logistic Regression with Complex Survey Data 143 6:4 Testing for a Reduced Model and Adequate Model Fit 153 6:5 Computing Unit-Level Statistics 156 6:6 Customizing Odds Ratios 157 6:7 Extensions for Modeling Variables with More than Two Outcomes 162 6:7:1 Introduction 162 6:7:2 Multinomial Logistic Regression Model for Nominal Outcomes 162 6:7:3 Cumulative Logistic Regression Model for Ordinal Outcomes 167 6:8 Summary 169 7: Survival Analysis with Complex Survey Data 172 7:1 Introduction 172 7:2 Foundations of Survival Analysis 173 7:2:1 Data Collection Strategies 173 7:2:2 Censoring 173 7:2:3 Definitions 176 7:2:4 Classification of Survival Analysis Models 177 7:3 Survival Analysis with Complex Survey Data 178 7:3:1 Visualizing the Data Using PROC LIFETEST 178 7:3:2 Fitting Cox Proportional Hazards Regression Models Using PROC SURVEYPHREG 183 7:3:3 Fitting Discrete-Time Hazards Regression Models Using PROC SURVEYLOGISTIC 192 7:4 Summary 200 8: Domain Estimation 202 8:1 Introduction 202 8:2 Definitions and an Example Data Set 203 8:3 The Risk in Subsetting a Complex Survey Data Set 205 8:4 Domain Estimation Using Domain-Specific Weights 210 8:5 Domain Estimation for Alternative Statistics 213 8:6 Significance Testing for Domain Mean Differences 219 8:7 Degrees of Freedom Adjustments 229 8:8 Summary 232 9: Replication Techniques for Variance Estimation 234 9:1 Introduction 234 9:2 More Details Regarding Taylor Series Linearization 235 9:3 Balanced Repeated Replication 238 9:4 Fay’s Variant to BRR 242 9:5 The Jackknife 245 9:6 The Bootstrap 250 9:7 Replication with Linear Models 254 9:8 Replication as a Tool for Estimating Variances of Complex Point Estimates 258 9:9 Degrees of Freedom Adjustments 262 9:10 Summary 264 10: Weight Adjustment Methods 266 10:1 Introduction 266 10:2 Definitions and Missing Data Assumptions 267 10:3 Adjustment Cell Method 273 10:4 Propensity Cell Method 278 10:5 Poststratification 280 10:6 Raking 284 10:7 Summary 288 11: Imputation Methods 290 11:1 Introduction 290 11:2 Definitions and a Brief Taxonomy of Imputation Techniques 291 11:3 Multiple Imputation as a Way to Incorporate Missing Data Uncertainty 294 11:4 Univariate Missingness 297 11:4:1 Introduction 297 11:4:2 Methods Based on Explicit Models 298 11:4:3 Methods Based on Implicit Models 302 11:4:4 A Semiparametric Method 307 11:5 Multivariate Missingness 308 11:5:1 Introduction 308 11:5:2 Methods for Monotone Missingness Patterns 309 11:5:3 Methods for Arbitrary Missingness Patterns 311 11:6 Inferences from Multiply Imputed Data 313 11:6:1 Introduction 313 11:6:2 Univariate Inferences 314 11:6:3 Multivariate Inferences 316 11:7 Accounting for Features of the Complex Survey Data during the Imputation Modeling and Analysis 320 11:8 Summary 322 References 324 Index 334 Complex Survey Data Analysis with SAS® is an invaluable resource for applied researchers analyzing data generated from a sample design involving any combination of stratification, clustering, unequal weights, or finite population correction factors. After clearly explaining how the presence of these features can invalidate the assumptions underlying most traditional statistical techniques, this book equips readers with the knowledge to confidently account for them during the estimation and inference process by employing the SURVEY family of SAS/STAT® procedures. The book offers comprehensive coverage of the most essential topics, including: Drawing random samples / Descriptive statistics for continuous and categorical variables / Fitting and interpreting linear and logistic regression models / Survival analysis / Domain estimation / Replication variance estimation methods / Weight adjustment and imputation methods for handling missing data. The easy-to-follow examples are drawn from real-world survey data sets spanning multiple disciplines, all of which can be downloaded for free along with syntax files from the author's website: http://mason.gmu.edu/~tlewis18/. While other books may touch on some of the same issues and nuances of complex survey data analysis, none features SAS exclusively and as exhaustively. Another unique aspect of this book is its abundance of handy workarounds for certain techniques not yet supported as of SAS Version 9.4, such as the ratio estimator for a total and the bootstrap for variance estimation.-- Provided by Publisher Complex Survey Data Analysis with SAS (R) is an invaluable resource for applied researchers analyzing data generated from a sample design involving any combination of stratification, clustering, unequal weights, or finite population correction factors.