Comparing Matlab and Apache statistics - kurtosis -

hi comparing statistics between matlab , apache functions. here apache functions tested in java. same set of data, different results double array (double[] ) follow:

---------------------------------------            matlab       vs apache --------------------------------------- max      = 0.5451       vs 0.5450980392156862 min      = 0.4941       vs 0.49411764705882355 var      = 5.4154e-05   vs 5.415357603461868e-5 std      = 0.0074       vs 0.007358911334879547 mean     = 0.5206       vs 0.5205525290240967 kurtosis = 3.3442       vs 0.35227427833465486 skewness = 0.2643       vs 0.26466432504210746 

i checked , rechecked data, each value matlabs same used in java. here can see statistics identical, except kurtosis.

is possible kurtosis computed differently matlab , apache library? if so, data should trust then?

---

edit

my data subset of image matrix (containing pixels values). each subset compute above statistics. everytime, statistics match except kurtosis.

the matlab code computing kurtosis of subset following:

kurtosis( sub(:) ); // sub n x m matrix 

while 1 used in java is:

import org.apache.commons.math3.stat.descriptive.moment.kurtosis; // ... kurtosis kurt = new kurtosis(); system.out.println("-kurtosis: " + kurt.evaluate(subimg) ); 

subimg being double[n x m] array.

you can calculate apache java statistics in matlab importing function. apache function uses unbiased estimator of population excess kurtosis. excess kurtosis means substracting 3 kurtosis of normal distribution equal zero.

to demonstrate made matlab function out of function (apache documentation):

function y = kurtosis_apache(x)      n=length(x);     mean_x = mean(x);     std_x = std(x);      y = ( (n*(n+1) / ((n -1)*(n - 2)*(n-3))) * sum((x - mean_x).^4) / std_x.^4 ) - ((3*(n-1).^2) / ((n-2)*(n-3))); end 

and code in command window shows matlab apache implementation, java apache implementation, , matlab version (biassed/unbiassed):

javaaddpath('commons-math3-3.2.jar') import org.apache.commons.math3.stat.descriptive.moment.kurtosis;  x = randn(1e4,1);  kurtosis_apache(x) 

ans = 0.0016

kurt = kurtosis(); kurt.evaluate(x) 

ans = 0.0016

kurtosis(x) 

ans = 3.0010

kurtosis(x,0) 

ans = 3.0016

note matlab kurtosis documentation:

so 0 flag unbiassed matlab implementation same apache version, when substract 3 make excess kurtosis.

(kurtosis(x,0)-3)-kurt.evaluate(x) 

ans = 3.8636e-14