Abstract: In this paper, the internal structure of the Mixing Data Sampling model(MIDAS)was thoroughly analyzed, and the difference and connection between the MIDAS model and the traditional EQW model were deduced, and the error of parameter estimation of EQW model was proved. On the basis of the above analysis, five different forms of weight function were constructed, and the concrete mechanism and deduction process of the nonlinear least squares estimation method of MIDAS model were obtained. Furthermore, the intra-sample forecast of the effect of asset price on China's economic growth was taken as an example to test the accuracy of MIDAS model. The empirical results show that the fitting effect and the intra-sample prediction accuracy of the multivariate EQW model are lower than those of the ExpAlmon-AR-M-MIDAS model under the optimal lag order; the ExpAlmon-AR-M-MIDAS model can extract more high-frequency interpretations.