Efficient Estimation in a Semiparametric Autoregressive Model

This paper constructs efficient estimates of the parameter in the semiparametric auto-regression model,with a smooth function and independent and identically distributed innovations _t with zero means and finite variances. This will be done under the assumptions that and that the errors have a density with finite Fisher information for location. The former condition guarantees that the process can be chosen to be stationary and ergodic.