Missing values: The observations missing from the set of data for some of the reason. In longitudinal studies, for instance, they might occur because subjects drop out of the study completely or do not appear for one or other of scheduled visits or because of the equipment failure. The common causes of subjects prematurely ceasing to participate include the recovery, lack of improvement, the unwanted signs or symptoms that might be related to the investigational treatment, unlikeable study procedures and the intercurrent health problems. Such values greatly complicate number of methods of analysis and simply using those individuals for whom data are complete can be unsatisfactory in number of situations. A distinction can be made between the values missing completely at random (MCAR), missing at random (MAR) and the non-ignorable (or informative).

The MCAR variety arise when the individuals drop out of study in a process which is independent of the observed measurements and those that would have been available had they not been missing both; here the observed values effectively constitute the simple random sample of the values for all study subjects. Random drop-out (MAR) happens when the dropout process depends on the outcomes which have been observed in the past, but given this information is conditionally independent of all future (which is unrecorded) values of the outcome variable following the drop-out. At last, in the case of informative drop-out, the drop-out process depends upon the unobserved values of the result variable. It is the latter which cause most the problems for the analysis of data comprising missing values.