Read more
Data fusion or statistical file matching techniques merge data sets from different survey samples to solve the problem that exists when no single file contains all the variables of interest. Media agencies are merging television and purchasing data, statistical offices match tax information with income surveys. Many traditional applications are known but information about these procedures is often difficult to achieve. The author proposes the use of multiple imputation (MI) techniques using informative prior distributions to overcome the conditional independence assumption. By means of MI sensitivity of the unconditional association of the variables not jointy observed can be displayed. An application of the alternative approaches with real world data concludes the book.
List of contents
1.1 Statistical Matching - Problems and Perspectives.- 1.2 Record Linkage Versu s Statistical Matching.- 1.3 Statistical Matching as Nonresponse Phenomenon.- 1.4 Identification Problems Inherent in Statistical Matching.- 1.5 Outline of th e Book.- 1.6 Bibliographic and Software Notes.- Frequentist Theory of Statistical Matching.- 2.1 Introduction and Chapters Outline.- 2.2 The Matching Process.- 2.3 Properties of the Matching Process.- 2.4 Matching by Propensity Scores.- 2.5 Obj ectives of Statisti cal Matching.- 2.6 Some Illustrations.- 2.7 Concluding Remarks.- Practical Applications of Statistical Matching.- 3.1 Introduction and Chapters Outline.- 3.2 History of Statistical Matching Techniques.- 3.3 Overview of Traditional Approaches.- Alternative Approaches to Statistical Matching.- 4.1 Introduction and Chapters Outline.- 4.2 Some Basic Notation.- 4.3 Multiple Imputation Inference.- 4.4 Regression Imputation with Random Residuals.- 4.5 Noniterative Multivariate Imputation Procedure.- 4.6 Data Augmentation.- 4.7 Iterative Univariate Imputations by Chained Equ ations.- 4.8 Simulation Study - Multivariate Normal Data.- 4.9 Concluding Remarks.- Empirical Evaluation of Alternative Approaches.- 5.1 Introduction and Chapters Outline.- 5.2 Simulation Study Using Survey Data.- 5.3 Simulation Study Using Generated Data.- 5.4 Design of the Evaluation Study.- 5.5 Results Due to Alternative Approaches.- 5.6 Concluding Remarks.- Synopsis and Outlook.- 6.1 Synopsis.- 6.2 Outlook.- Some Technicalities.- Multivariate Normal Model Completely Observed.- Normally Distributed Data Not Jointly Observed.- Basic S-PLUS Routines.- EVALprio.- EVALd.- NIBAS.- Tables.- References.
Summary
Data fusion or statistical file matching techniques merge data sets from different survey samples to solve the problem that exists when no single file contains all the variables of interest. Media agencies are merging television and purchasing data, statistical offices match tax information with income surveys. Many traditional applications are known but information about these procedures is often difficult to achieve. The author proposes the use of multiple imputation (MI) techniques using informative prior distributions to overcome the conditional independence assumption. By means of MI sensitivity of the unconditional association of the variables not jointy observed can be displayed. An application of the alternative approaches with real world data concludes the book.
Additional text
"Statistical matching is one of the methods for the imputation of missing results from survey sampling…This book is totally focused on methods, applications, and results. It devotes little space to the presentation and derivation of various theorems." Technometrics, May 2004
Report
"Statistical matching is one of the methods for the imputation of missing results from survey sampling...This book is totally focused on methods, applications, and results. It devotes little space to the presentation and derivation of various theorems." Technometrics, May 2004
