Conditional Variance Calculation

[sirooskhadem]

Hi,

I have been trying to include conditional variance into my model. I have extracted the conditional variance (garch) from the return equation but I would like to know how eviews calculates it.
I have tried this formula [Var = (1/N)*(Return-(@mean(return)))^2]. graphically they are very close but the values are different. I would like to know how eviews calculates the conditional variance then base on that I would be able to also estimate the Conditional skewness and kurtosis.

I would be grateful if anyone can advise me.

Regards,
Siroos

[EViews Glenn]

I'm a little confused by the question. As one might expect, since you estimated a GARCH model, EViews uses your GARCH specification, not the unconditional variance, to compute the conditional variance. It is not surprising that the two differ.

[sirooskhadem]

Thank you very much, now it makes scense.

Now my questions is if the formula that I have applied for conditional variance is correct?
Var = (1/N)*(Return-(@mean(return)))^2

What if I remove the scaling by 1/N since I have more than 700,000 observations. Will it be better?

and finally how about the skewness and kurtosis, do they make scense? and if they are correct should I keep the (1/N) or not?

Skew = (1/N)*((Return-(@mean(return)))/(@stdev(return)))^3
Kurt = (1/N)*((Return-(@mean(return)))/(@stdev(return)))^4

Many thanks,
Siroos