摘要：We present a new heteroskedastic conditional variance model using Non-Linear Moving Average as the basis for this specification []. The typical problem of this class of models-i.e., non-invertibility—is solved by means of an intuitive parametric restriction; this allows us to use Maximum Likelihood as the estimation procedure. The statistical properties of the new model are both simple and attractive for empirical purposes in finance: a natural fat-tailed distribution stands out. The Autocorrelation Function of the squared process allows us for identification of the number of lags to be included in the new specification. In addition, we present several Monte Carlo experiments where the properties of the model using finite samples are exhibited. Finally, an empirical application using exchange rates and capital market bonds is shown. ()NLMACHq.
