Impact of Missing Data on Parameter Estimation Algorithm of Normal Distribution

This paper propose a simulation approach for the parameter estimation of normal distribution, to analyze the EM algorithm with missing data under different missing rates and complete data maximum likelihood estimation. The simulation result shows that the EM algorithm and the maximum likelihood estimates are almost unanimously when the missing rate is less than 0.25, but the effect of parameter estimation of the EM algorithm gradually deteriorates when the missing rate increases. The result also shows that the EM algorithm is more sensitive to the initial value. In addition, this paper also analyzes the evaluation of the selection of initial value for EM algorithm.