jamiembright / time-series-validation Goto Github PK

Hello,

Thanks for the code. I was looking at the original paper by Espinar et.al (10.1016/j.solener.2008.07.009) on calculation of KSI and understand that the variable 'Dn' is the absolute difference between the two CDFs (Observed and Predicted) at each interval. However, the code calculates 'Dn' as the absolute difference between the two frequency distributions. Shouldn't lines 263-266 be changed as follows?

Od = cumsum(histc(O,xbins)./N);
Pd = cumsum(histc(P,xbins)./N);
% absolute difference between the two CDFs
Dn = abs(Od - Pd);

And the parameter OVER is the integration of (Dn-Vc) over the irradiance spectrum, Where Vc=1.63/sqrt(N). The code however considers all the points that are above Vc for integration, while setting the values below Vc to zero.

Please let me know if my thinking is correct.

Many thanks