MISSING VALUES ANALYSIS & DATA IMPUTATION

Overview

Proper handling of missing values is important in all statistical analyses. Improper handling of missing values will distort analysis because, until proven otherwise, the researcher must assume that missing cases differ in analytically important ways from cases where values are present. That is, the problem with missing values is not so much reduced sample size as it is the possibility that the remaining data set is biased. The imputation of values where data are missing is an area of statistics which has developed much since the 1980s.

The SPSS add-on module "Missing Value Analysis" has long supported several imputation algorithms, the most popular being expectation maximization (EM). Since SPSS 17 a separate module, "Multiple Imputation," has supported the newer and increasingly preferred MI estimation method. Both are discussed below. Note that maximum likelihood data imputation, an EM method, can also be implemented in AMOS, the structural equation program supported by SPSS.

Statistical objections can be raised about any of the methods which might be used for data imputation. Missing data are a form of measurement error. As such missing data may both bias the sample and attenuate effect sizes. Data imputation may reduce bias but also may introduce systematic regularities in the data arising from the prediction method.

The full content is now available from Statistical Associates Publishers. Click here.

Below is the unformatted table of contents.

Table of Contents Overview 5 Key Concepts and Terms 5 Item non-response 5 Dropping cases with missing values 6 Types of Missingness 6 Missing completely at random (MCAR) 6 Missing at random (MAR) 8 Non-ignorable missingness 10 Multiple imputation (MI) in SPSS 11 Overview 11 MI vs. MVA 11 How MI works 11 What MI does 12 Using MI data in SPSS 13 SPSS multiple imputation set-up 15 The Method tab 17 The Constraints tab 18 The Output tab 19 Multiple imputation output in SPSS 20 The imputation models table 20 The descriptive statistics tables 20 Missing Value Analysis (MVA) in SPSS 21 What MVA does 21 MVA set-up in SPSS 22 Types of estimation 23 The variables button 25 The patterns button 26 The descriptives button 29 Other MVA output 30 Default output 30 The percent mismatch table 30 Output for t tests 31 Crosstabulation 34 Replacement of missing values 34 Mean substitution 35 Frequently Asked Questions 36 Why not just delete cases with missing values rather than impute values at all? 36 Should I use original data or imputed data when reporting results? 36 What is approximate Bayesian bootstrapping? 36 Bibliography 37