Abstract: Firstly, the  Bayesian method was used to study  the problem of parameter estimation in the NGINAR(1) model, and then the method of conditional distribution prediction was used to predict  the NGINAR(1) model. The  simulation result shows that Bayesian estimation is better than the conditional least squares and YuleWalker estimation, and the method of conditional distribution prediction is more suitable for integervalued series than the method of conditional expectation prediction.

Key words: NGINAR(1) model, Bayesian estimation, model prediction