Most data sets are affected by missing values, i.e., data values of scientifically interesting variables that are assumed to exist but are not observed and can not deterministically be derived from observed values. This characterization is not a clear-cut definition of the phenomenon of missing values, but it includes the most common situations. Examples are unanswered questions or not reported reactions of statistical units in general, values which are not observed because units are exposed only to parts or blocks of a larger questionnaire (‘missings by design'), if impossible or implausible values are deleted, or, in the context of causal inference and non-experimental settings, when units are observed only under one of two (or more) conditions. ...
Dealing with Missing Values
Dealing with missing values