Log-Range volatility for dummies?
- Thread starter mgdpublic
- Start date
It is an alternate method used to model volatility with results that are argumentatively closer to empirical data, and have a more Gaussian distribution
than prior methods. As I briefly understand it, in very simple terms it is the log of the high price minus the log of the low price over a given sample window.
paper:
http://fic.wharton.upenn.edu/fic/papers/00/0028.pdf
laymen's:
http://www.investment-analytics.com/files/Articles/Modeling Asset Volatility.pdf
I haven't run the numbers, personally, but I do wonder about these posters who pop in for one or two detailed questions to elicit quant responses, then disappear as if they really didn't have any interest to begin with.
I don't see what the big deal is.
log( x / y ) = log(x) - log(y)
You take the top price, divide it by the bottom price and take the log for an arbitrary fixed window size? Your max peak-to-trough return possible for a single hold, and that becomes a measure of the volatility. Fine, I guess. May be good for quick calculations, vs something more sophisticated.
As the window shifts, if a new peak enters the window it grows, otherwise it stays the same. If a new trough shows up, the vol goes up. Seems like any other technical indicator. Basically rejecting any move that isn't significant with respect to the current window. Kind of a lazy way to filter out the usual changes that aren't at the tails.
log( x / y ) = log(x) - log(y)
You take the top price, divide it by the bottom price and take the log for an arbitrary fixed window size? Your max peak-to-trough return possible for a single hold, and that becomes a measure of the volatility. Fine, I guess. May be good for quick calculations, vs something more sophisticated.
As the window shifts, if a new peak enters the window it grows, otherwise it stays the same. If a new trough shows up, the vol goes up. Seems like any other technical indicator. Basically rejecting any move that isn't significant with respect to the current window. Kind of a lazy way to filter out the usual changes that aren't at the tails.
Quote from mgdpublic:
You've unearthed our cabal of drive-by forum posters!
Perhaps if you took the time to read the rest of the sources I conveniently summarized and posted for you, you might get the answer that you are seeking.
As I recall, you did ask for a simplified summary, right?
So, cabal denotes a bad connotation.
But now that you've opened Pandora's box,
how about preferably listing your ahem, aliases and other affiliated one liner, seemingly interested query posters? Do tell.
Some of us are fascinated by information gathering and behavioral response post-count generating methodologies.
As I understand from a quick look is that they instead of looking at the log of close to close returns use the log of the high to low ratio as the "return" in the calculation of the volatility estimate. So its not just the high to low over the entire window which, as you say, would become a static measure until a new high/low is encountered in the window.
For days with a large high to low ratio but where we end up with the same close price as yesterday this would be a better estimate of "true" volatility.