Abstract: Firstly, by using the method of conditional least squares, we discussed the problem of parameter estimation for MGINAR(p) model with missing data, and obtained the conditional least square estimations of parameters. Secondly, numerical simulation was used to verify the feasibility of four methods for processing missing data and to compare the estimation effects. The simulation results show that when the missing probability is small, the case rejection method or the mean value interpolation method can be used, and when the missing probability is large, the bridge interpolation method can be used to reduce the estimation bias.

Key words:  , MGINAR(p) model, missing data, parameter estimation