Volatility: Moving Average Approaches

[waldymar100]

apologize for not being clear. i meant from your video "how to scale autocorrelated returns" you are talking about autocorrelation among returns and pointing that in presence of autocorrelation we should use "ar(1) scale factor" in order to calculate the "true" volatility. so, my question here is: since we are talking about exchange rate and it's a time-series data, should we scale it as well to get the "true" volatility? in your case, should cell F42 be multiplied by ar(1) scale factor?

[waldymar100]

Hi David,

should we consider AR(1) adjustment factor (topic you have in you arhive) to correct the variance you found? as you know time series are correlated, therefore the volatility will be higher than the simple standard deviation.

[bionicturtledotcom]

I don't follow you. It's just the calc of a daily standard deviation; no adjust required. If we scale it to annual standard deviation (is that what you mean by volatitlity?), then the scaling may be adjusted either way (for autocorrelation or mean reversion)...but no adjustment needed to the historical period Std Deviation - David

[tubehole08]

Why are you using a natural log of Si/Si-1? Why not just [(Si/Si-1)-1] ?

[vincentlcarter]

@tubehole08 natural log is a more accurate measure (though i think u should not notice much difference between the 2 methods)