[Gretl-users] Re: GARCH News Impact Curve

Am 13.07.2022 um 15:06 schrieb Riccardo (Jack) Lucchetti: Not sure if I'm missing something, but the concept of the NIC doesn't seem so hard, so I'm attaching an attempt of a function that provides the ingredients for the plot. Please correct me if I misunderstood what's wanted. Notice that I'm neglecting all ARCH orders beyond 1 because the e^2 in the more distant past are supposed to be zero in the hypothetical scenario of a shock only occurring in t-1. (As opposed to the GARCH higher orders, which do contribute.) I hope that's right. The function only calculates values on a simple discrete grid which can then be plotted. I'm sure Jack or somebody else could also cook up some gnuplot code that takes the algebraic quadratic function directly and plots the theoretical continuous line. (The parameters are returned by the interim function NICparams.) Implementing this thing was complicated a little by the fact that the $model accessor after 'garch' apparently doesn't store the specified p and q orders, so I had to infer these from the parnames contents (in the getGARCHorders helper function). It works, but it would be nice if p and q could be added to the model bundle in the future. Finally, so far this only works for the native GARCH (including the ARCH nested case), not for the stuff in the gig package. cheers sven