Evaluation of Imputation Techniques for Genotypic Data of Soybean Crop under Missing Completely at Random Mechanism

Background: The issue of missing data is prevalent in all type of research work, which can diminish statistical power and lead to inaccurate results if not managed correctly. Missing data cannot be ignored because every piece of data, no matter how small, affects the outcome significantly. Imputation is a key component in dealing with missing data; however, the best way to impute missing values has not yet been identified. Methods: Our goal of this paper is to compare four more recently developed imputation techniques - MICE, MI, missForest and Amelia. In order to examine the performance of various imputation techniques, non-missing data were deleted from genotypic data of soybean crop with varied frequency of missingness by missing completely at random mechanism. The study compared different imputation techniques for solving missing values using the root mean square error and mean absolute error. Result: To fill in the dataset’s missing values, the imputation technique producing the lowest value of the RMSE and MAE will be taken into consideration. Finally it is observed that missForest technique performs best on the genotypic data of soybean at different proportion of missingness.