Hurst Exponent-Rescaled Range Analysis

[trubador]

Please find attached two subroutines that each provides a different estimation of Hurst Exponent for Rescaled Range Analysis.

Code: Select all

subroutine hurst(series y) series ym !n = @obs(y) series RS_sum series RS_ave for !i=2 to !n !count = 1 series R for !j = 1 to !n-!i+1 step !i series cumser smpl !j !i+!j-1 !S = @stdev(y) series ym = y - @mean(y) cumser = @cumsum(ym) R(!count) = (@max(cumser) - @min(cumser))/!S !count = !count+1 next smpl 1 !count-1 RS_sum = @cumsum(R) RS_ave(!i) = RS_sum(!count-1)/(!count-1) next smpl @all !start = @floor(10^(@log10(!n)*.25)) !end = @floor(10^(@log10(!n)*.75)) smpl !start !end equation eq1.ls @log10(RS_ave) c @log10(@trend(0)) vector(1) hurst = eq1.@coefs(2) smpl @all delete cumser r rs_ave rs_sum ym endsub

Code: Select all

subroutine hurst(series y) series ym series R !n = @obs(y) for !i=2 to !n !R_sum = 0 !S_sum = 0 for !j = 1 to !n-!i+1 series cumser smpl !j !j+!i-1 !S = @stdevp(y) ym = y - @mean(y) cumser = @cumsum(ym) !R = @max(cumser) - @min(cumser) !R_sum = !R_sum + !R !S_sum = !S_sum + !S next R(!i) = !R_sum / !S_sum next smpl @first+2 @last equation eq1.ls @log10(R) c @log10(@trend(0)) vector(1) hurst = eq1.@coefs(2) smpl @all delete ym r cumser endsub

[mmc23]

hi Mr trubador
I have simple suggestion, would you like to write an add-in for test the long memory which includes (R/S classic and Lo's, Moody and Wu test, GPH, GSP test...)
i have seen that the other software like Rats and stata matlab can do this test, if possible also the add-in can calculate hurst with many methode like R/S and Rescaled Variance...etc (Ref 'Long-Memory Processes Probabilistic Properties and Statistical Methods, Springer 2013, Jan Beran & al, pp 409-1-416').

best regards.

[ahmedex]

This short program was posted last year (actually few years back). I found it very useful to compute Hurst coefficient. However, I wonder if the original author of this program or someone else can post additional commands to obtain R (square) and F-statistics from the series which generates Hurst coefficients.