ABSTRACT
This chapter considers semiparametric modelling of nonlinear time series data. We first propose an additive partially linear modelling method. A semiparametric single-index modelling procedure is then considered. Both new estimation methods and implementation procedures are discussed in some detail. The main ideas are to use either a partially linear form or a semiparametric single-index form to approximate the conditional mean function rather than directly assuming that the true conditional mean function is of either a partially linear form or a semiparametric single-index form.
