In this paper we suggest new method to estimate the missing data in bivariate normal distribution and compare it with Single Imputation method (Unconditional mean and Conditional mean) by using simulation.