[R-SIG-Finance] How to estimate the Hurst exponent in a stable and effective way?

[Mogei Wang]

Hello, very one,
I am using the package "fArma" to compute the Hurst exponent.
I have tried all the 9 *Fit functions, and got quite different results
from different functions.
And sometimes, the result is larger than 1, or less than 0.

In the program, I just call the fit functions
*Fit(myData)@Hurst$H
,       where myData is a vector (like c(1,2,3,4,5)) with a few hundreds
      elements (usually 200 to 1000).
Is this the right way to call these *Fit functions?
And are there parameters that must be carefully set?

In short, how to estimate the Hurst exponent in a stable and effective way?

Thanks ahead!

version of R: 2.10.1
version of fArma: 2100.76.

Yours, M.W.

The following is a little program for testing the *Fit functions:

----------------------------
require(stats);
require(graphics);
require(fArma);

myData <- sin((0:500)*1);
plot(myData,type="b");

print(list(pengFit(myData)@hurst$H,
rsFit(myData)@hurst$H,
aggvarFit(myData)@hurst$H,
higuchiFit(myData)@hurst$H,
absvalFit(myData)@hurst$H,
diffvarFit(myData)@hurst$H,
perFit(myData)@hurst$H,
boxperFit(myData)@hurst$H,
whittleFit(myData)@hurst$H));

----------------------------
The results of the *Fit functions:
0.01589469
0.1455046
0.1131107
-0.02142119
0.162419
0.1993037
0.358259
0.6157515
0.9951347