Semiparametric time series models driven by latent factor

We introduce a class of semiparametric time series models by assuming quasi-likelihood approach driven latent factor process. More specifically, given the process, we only specify conditional mean and variance enjoy function for estimating parameters related to mean. This proposed methodology has three remarkable features: (i) no parametric form is assumed distribution process; (ii) able modelling non-negative, count, bounded/binary real-valued series; (iii) dispersion parameter not be known. Further, obtain explicit expressions marginal moments autocorrelation process so that method can employed also Simulated results aiming check estimation procedure are presented. Real data analysis on unemployment rate precipitation illustrate potencial practice our methodology.