Missing Data Imputation

In clinical research, an instrument (or questionnaire) is often used to provide a standardized and objective means of collecting data on subjective states or events across various therapeutic areas. In practice, although there exist many instruments such as Hamilton-D (Hamilton scale for depression) and Hamilton-A (Hamilton scale for anxiety) for central nervous system (CNS) and quality of life (QOL) assessment in cancer trials, the investigators frequently face the need to develop new ones. This need arises because a proper development and validation of the existing instruments was achieved for a speciŒc purpose and relative to a speciŒc target patient population. While the existing pool of instruments may contain one that has been developed for the target patient population and the desired purpose, new research questions often require new instruments for measurement. Validation of the developed instrument is important to ensure a proper sampling and a valid measurement of the content of the subjective state, behavior, or disease to be measured (Testa, 1987). For illustration purpose, in this chapter we will focus on the validation of QOL instruments. The performance characteristics for the validation of QOL instruments can be applied to other instruments for other purposes across therapeutic areas.