A Comparison between (ECM) and (KNN) Methods for the Multivariate skew-normal model with incomplete data

Abstract

Statistical parameter estimation for multivariate data leads to waste of information, if the missing data are omitted, and in return will lead to inaccurate estimates, that is why incomplete data should be estimated using one of the statistical estimation methods to get accurate results and in return good estimates for the parameters. This paper aims to estimate the missing values for a multivariate skew normal model using the Expectation Conditional Maximization (ECM) algorithm and the K-Nearest Neighbour (KNN) method. After estimating the missing values, the model parameters will be estimated using the Maximum Likelihood Estimation (MLE) method and Newton-Raphson algorithm. A comparison between the (ECM) algorithm and the (KNN) method was conducted using the Mean Squared Error (MSE) for the model through simulation to find the best estimation method.