A SURVEY ON MISSING DATA AND METHODS TO FIND THE MISSING VALUES

Abstract

Missing data plagues almost all surveys, and quite a number of designed experiments. No matter howcarefully an investigator tries to have all questions fully responded to in a survey, or how well designedan experiment is; examples of how  his can occur are when a question is unanswered in a survey, or aflood has removed a crop planted close to a river. The problem is, how to deal with missing data, once ithas been deemed impossible to recover the actual missing values. Traditional approaches include case deletion and mean imputation. These are thedefault for the major Statistical packages. In the last decade interest has centered on RegressionImputation, and Imputation of values using the ExpectationMaximization algorithm, both ofwhich will perform Single Imputation. More recently Multiple Imputation has become available, and isnow being included as an option in the mainstream packages.