The effects of completion algorithms on parameter estimates in normal, multinomial, exponential and gamma distributed data structures and missing data types (MCAR, MAR, MNAR)

The missing data problem is one of the main problems that researchers constantly encounter when performing statistical analysis. Missing data may result from the individual conducting the work, the inability of the data to be collected, the inability to observe the data, the equipment used to record the data, and the like. The biggest problem with the lack of data is the need for a complete set of data for almost all statistical analyses. In this study, data were produced according to four types of distribution, these data were decreased in certain ratios according to MCAR, MAR and MNAR types which are incomplete data mechanisms and these minimized data were completed according to missing data analysis methods. Calculated data, deducted data and parameters of the completed data were calculated and compared with these values. As a result, it has been determined that multiple imputation method makes more effective and neutral parameter estimation compared to other methods.