Imputation And Deletion Methods Under The Presence Of Missing Values And Outliers: A Comparative Study

Missing data and imputation methods are studied in many disciplines. However, the methods have some different properties and some constraints according to missingness mechanism. In this paper, we examine some deletion and imputation methods' behaviors under the presence of outliers. We obtain a mean vector and covariance matrix with missing and contaminated data and compare the results of imputation methods using mean square errors. As an application, we use the regression data and examine the effect of missingness on regression model's parameters. We compare the imputed values with real values and explain the results of classical and robust imputation methods.