Abstract: The parameters of a stationary NEAR(p) (new exponential autoregressive process of order p) model were estimated by means of conditional least squares, weighted conditional least squares and maximum quasi likelihood approach, respectively. For those estimators, we discussed their asymptotic properties and showed via simulation that maximum quasilikelihood es timation dominates the others when the true values of the parameters are small, while weighted conditional least squares estimation dominates the others when the true values of the parameters are big.

Key words: NEAR(p) model, conditional least squares estimati on, weighted conditional least squares estimation, maximum quasilikelihood est imation