Hurst exponent

["Walter Lake" <wlake@xxxxxxxxx>]

Thanks for your emails

You probably noticed the similarity between the following paragraphs at

http://www.speedderivs.com/hurst.htm

"... Where "H" is the Hurst exponent and "c" is a constant. Linear
regression is performed on "ln((R/S)sub.N)" and "ln(N)". The gradient of the
regression line is the Hurst exponent. ..."

and at

http://oara.org/mpc/fma/fmatutr2.htm

"... When you compare the powers of two related numbers (such as the log of
the number of observations and the log of the rescaled range for each of
those numbers) you get a power scale.

That's what Rescaled Range analysis does. As the scale increases
logarithmically for the number of observations, we would expect the range of
R/S values among those observations to scale accordingly. The log/log plots
of R/S measure this power relationship. Specifically, the slope of the
regression line that plots the power scale is an estimate of the Hurst
exponent (H). ..."

Best regards

Walter