ABSTRACT
Traditional use of imputation was motivated by the desire to avoid bias due to missing data and the unavailability of analytic methods for incomplete data. Prior to the mid-1980s, statistical software for longitudinal studies or repeated measures, such as MANOVA or MANCOVA, required the exclusion of all individuals with any missing data. This disadvantage has disappeared with the accessibility of software for incomplete data such as maximum likelihood estimation (MLE) for longitudinal studies with ignorable missing data [109, 115, 132] described in Chapter 5. When it is reasonable to assume that the missing data in a HRQoL study are either MCAR or MAR, these MLE procedures are appropriate and the need to impute missing observations diminishes.
