In this paper the problem of estimating the population mean by using information of an auxiliary variable in presence of non-response has been considered. For this purpose some ratio, product and regression estimators for estimating the population mean in two-phase sampling have been suggested and their properties have been studied. These estimators are generally biased and the theoretical comparisons of their mean square errors are difficult. So by neglecting biases of these types of estimators and comparison of variances, the best estimator of population mean has been derived. By a simulation study, effects of different values of correlation coefficient on mean square error of these estimators are compared for different values rate of non-response and size of sample. Results of this simulation study show that the simulated mean square will reduce when the correlation coefficient between auxiliary and study variables is increase, rate of non-response is decrease and the size of sample is increase.