Statistical Analysis with Missing Data., 3rd Edition

3Complete-Case and Available-Case Analysis, Including Weighting Methods

3.1 Introduction

In Chapter 2, we discussed the analysis of data with missing values confined to a single outcome variable that is related to completely observed predictor variables through a linear model. We now discuss the more general problem with values missing for more than one variable. In this chapter, we discuss “complete-case” (CC) analysis, which confines the analysis to the set of units with no missing values and modifications and extensions. In the following two chapters, we discuss imputation methods. Afifi and Elashoff (1966) review the earlier literature on missing data, including some of the methods discussed here. Although these methods appear in statistical computing software and are still widely used, we do not generally recommend any of them except in situations where the amount of additional missing information in the incomplete units is limited. The procedures in Part II provide sounder solutions in more general circumstances.

3.2 Complete-Case Analysis

CC analysis confines attention to cases (units) where all the variables are present. Advantages of this approach are (i) simplicity because standard complete-data statistical analyses can be applied without modifications and (ii) comparability of univariate statistics because these are all calculated on a common sample base of units. If the additional information for the target parameters in the incomplete units is small, then including ...