Roman Pavelka, Statistical Office of the Slovak Republic, Slovak Republic

Type of article: scientific article
Pages: 41 – 65

Abstract

Missing values represent a complication in most statistical analyses. Observations with incomplete values of variables (also called incomplete cases) are implicitly excluded from statistical analyses of the detected observations in most statistical softwares. The use of complete cases (observations with completely filled values for all variables), is simple, though usually redeemed by the loss of information due to the exclusion of incomplete cases. Moreover, excluding incomplete observations from statistical analyses also ignores potential systematic differences between estimates and actual values, and the resulting statistical inference may not be applicable to the population of interest for all statistical units (all cases), especially in conditions of fewer complete cases. For this reason, it is therefore important to analyse not only the observed data, but also to recognise the ‘missing’ mechanism (statistical model or probability distribution) of incomplete data in order to appropriately fill them with acceptable (imputed) values. Although there are many different methods for working with incomplete data, the multiple imputation method has become one of the most important methods for dealing with the incompleteness of the observed data. Even in the SAS analytical system environment, the use of the multiple imputation method represents one of the options for dealing with incomplete data, and this is the content of the present paper.