|
Abstract:
|
A generalized autoregressive score, or GAS, model is a general framework within which a time series with time-varying parameters can be modeled (e.g., GARCH). This framework uses the score function, derived from the conditional density function of the observed series, to update the time-varying parameter value at each time period. However, in cases where this conditional function has no closed form, is complex, etc., the score function may be difficult or even impossible to derive. We show how an optimal estimation function (i.e., quasi-score) can be used in place of the score function to handle this issue.
|
