Serving the Quantitative Finance Community
the Mean
would using the following GARCH type specification for the conditional mean make any sense (assuming p=1 & q=1):r(t)=alpha0+alpha1*(R(t-1)-r(t-1))+alpha2*sigma(t-1)for daily index time series for getting the normalising parameters in ((xi-r(t)/sigma)? or would a simple rolling mean be better, or even the unconditional mean, or something else? or is this complete bs?rgds,.k.
the Mean
QuoteOriginally posted by: SPAAGGIt depends on the market, the asset class, whether you backtest or you forcast, which is your mesure to determine the quality of your fit.... --> not able to answer...sorry, equity market indices for developed markets (US, Europe, Japan) and developing markets, essentially backtesting. i don't really need the mean for anything else but normalising, but i need a sensible process for it, so the measure could simply be RMSE/MAE, if that's what you meant by the measure? so does it make sense?