this is slightly incorrect. GARCH(p,q) is a generalization of the ARCH(p) model since it allows for an expression the conditional variance in terms of both auto-regressive terms up to order p and moving average terms up to order q. EWMA is simply a moving average model with the key distinction that the model defines an exponential decay that persists for very high orders, yet only the decay rate is specified.
great point. For moving average (unweighted - left side above), order of historical returns does *not* matter. But that is it's problem: price trend up or down can give the same volatility. For the other approaches (ARCH(m)), order matters because more recent returns are given more weight....I can't tell be looking at graph, maybe some could..
David Harper, if you ever come to Belgium, I will buy you a beer.
viavo 2 weeks ago
@viavo I wrote that down, Belgium!
bionicturtledotcom 2 weeks ago
this is slightly incorrect. GARCH(p,q) is a generalization of the ARCH(p) model since it allows for an expression the conditional variance in terms of both auto-regressive terms up to order p and moving average terms up to order q. EWMA is simply a moving average model with the key distinction that the model defines an exponential decay that persists for very high orders, yet only the decay rate is specified.
oligiscool1 1 year ago
great point. For moving average (unweighted - left side above), order of historical returns does *not* matter. But that is it's problem: price trend up or down can give the same volatility. For the other approaches (ARCH(m)), order matters because more recent returns are given more weight....I can't tell be looking at graph, maybe some could..
bionicturtledotcom 4 years ago
quick questions on volatility calculation: does the calucalted value of the volatility depends on the order of the input variable?
is volatility always used +ve?
can one tell the volatility value just from looking at a graph?
does the input have to be +ve?
lilacfun 4 years ago