[R-SIG-Finance] GARCH Modelling of transformed series

[Бобровский Дмитрий]

 Hello. I have a really stupid question but... I break my head trying to understand, why fitted quantiles are so poor for my model.
Ok, let me tell about this problem.

Let's have a series (of stock data) - X_{t}. Transform it to Y_{t} = (X_{t} - Mean(X_{t}) / Val, where Val - same constant, which not connected to the series. 
Suppose, that Y_{t} stationary, and fit arma(p,q)+garch(P,Q) model for Y_{t}. So, is it correct to get quantiles for X_{t} as "fitted mean + quantile(conf_lvl, sd = 1, same_distrib) * sqrt(cond.var)" (here same_distrib like distrib used to fit garch part)?

Sorry, I know, this is really stupid question, but I need of your little help...

Thank you.

Best regards,
                         Dmitriy.


-- 
Дмитрий Бобровский
	[[alternative HTML version deleted]]