Hi everyone,I'm trying to do some multiperiod forecasting with an egarch model. Tsay explains in his book "Analysis of Financial Time Series" a method which relies upon the distribution used for fitting the model. But Zivot mentioned in an article ("Practical Issues in univariate GARCH modeling, in "Handbbook of financial time series") that a recursive method (which would not be directly related to underlying distribution) is also possible. I'm not sure about this becasue as far as I understood Tsay a simple recursion is not possible...Does anyone have any clue about this? Hints to related paper or other books would be really fine!thx a lot

A simple GARCH(1,1) can be recursively forecast by computing E[sig^2(T)] from the defn, right?So, presumably, in EGARCH, you can only easily forecast E[log sig^2(T)]. But, if instead you want to computeE[sig^2(T)] in EGARCH, I would just simulate the process using an easy Monte Carlo. BTW -- no need to double post.