Theory and Application of Two (2) Iterative Imputation Approaches to Nigeria Annual Rainfall Data Reported

Abstract

This research work is based on missing data statistics. Missing data occur where one or more of the observations in a dataset are completely not available. This work focuses on two (2) iterative imputation approaches. These are the Regression approach and the Expectation Maximization iterative imputation. These approaches were used to analyze the secondary data of the thirty-six (36) states in Nigeria on the rainfall data collected from the Annual Abstract of Statistics 2016. The evaluation criteria and comparison of these two approaches were done based on the error efficiency using the Raw Bias (RB), Mean Squared Error (MSE), Root Mean Squared Error (RMSE) and variance. The analysis of the result showed that the Expectation Maximization (EM) method was better for this specific data as reported in the Annual Abstract of Statistics 2016, compared to the other approaches. This was seen in the smaller errors values from the computed cases. It is therefore recommended that this approach should be used for obtaining missing data like other rainfall data in Nigeria. These two imputation approaches are good for making available missing data in observations.