How i can estimate the hurst parameter for a matrix of topographic data?

%%Author: Friedrich Leisch
%%Modified by Alois Schloegl

function H = hurst (x)

    if (nargin ~= 1)
        fprintf(1,'usage: hurst (x)');
    end

    if (isscalar (x))
      error ('hurst: x must not be a scalar')
    elseif (isvector (x))
      x = reshape (x, length (x), 1);
    end

    [xr, xc] = size (x);

    x0 = center(x,1);

    xr = sum(~isnan(x0)); 
    x0(isnan(x0))=0;

    s = sqrt(sum(x0.*x0,1)./max(0,xr-1));  % std
    w = cumsum (x0,1);
    RS = (max(w) - min(w)) ./ s;
    H = log (RS) ./ log (xr);

i found this code but don't works in matlab, it comes from a similar code from Octave

Answer 1

[Walter Roberson]

I'm having difficulty finding a reference page for exactly what Octave's center() function is intended to do.

It appears that possibly you could replace

x0 = center(x,1);

by

x0 = bsxfun(@minus,x,mean(x));

[Matt Fig]

Assuming the CENTER function was not part of a bundle that came with the HURST function...

Answer 2

[Matt Fig]

Would you mind elaborating on what "It doesn't work" means? Does it error? If so, what does the error message say. If not, what is the problem? If it is just completely wrong, why post it?? Did this function come as part of a bundle which included the CENTER function?

[Walter Roberson]

Octave does have a center() function; I have been able to find an official patch updating it. I just haven't been able to find documentation for it; in particular I am missing information about what the optional parameter 1 means in this situation.

[Walter Roberson]

In Octave it is statistics/center.m and in general terms it outputs x - E[x] where E[x] denotes the Expected Value of x. But I don't know whether it is intended to be per-column, or if it defaults to the expected value of the entire matrix but the "1" tells it to operate per-column, or if the "1" is acting as a redundant specification that it should act along the first dimension, or ... ?