Hi all,I'm not a statician.. but if I want to compute the effect of the "currency factor" or the "country economic growth rate", or the "P/E factor", or the lagged return of the stock, on a stock return (and risk), I think I'll ve to do a regression model.If I don't want to build a random N(0,1) time series return for the white noise, how can I compute the disturb?thanx a lot

Can you please rephrase this? What is it that you’re trying to do? My feeling is that you have a model (linear?) for one of the economic factors, and you want to see the possible paths (set of paths) this model might produce, as the future conditions are represented with some ‘noisy’ factors approximating in behavior what we see in the real world?

Hi Gee,Sorry..now I explain my problems better, I hope.At first, my aim's to predict future stock price by a linear (lagged) model (AR(p)).I use the daily return time series and 2 lags.My problem's about white noise component: it's a casual number from a N(0,1).I've liked to do a more "realistic" thing, so: r(t) = a + b* r(t-1) + c* r(t-2) + e(t-1)where e(t-1) it's the last regression error.It seems the model's not too bad, but I'd like to try it on weekly data.I studied these models some years ago and I know my model isn't correct...(I use an OLS on daily data and I haven't looked at ACF , for example). So I'm asking for some advice to make it working better (if possible..).So, is there a way to build the error by using fundamental data (forexample) ?I looked something as P/E ratio, volume, but I don't know how I could use itand if they could help me.