A serious consequence of missing data is the possibility of bias in the results. A variety of statistical approaches have been proposed with the objective of determining the missing data mechanism (MCAR, MAR, MNAR) and making suitable analyses to compensate for this. Two broad approaches are available. The first is to adapt the method of analysis to accommodate the missing data. The second is to "impute" (estimate) what the missing values are most likely to have been, and then apply standard methods of analysis or more sophisticated methods. The main approaches will be described and illustrated.