How to compare statistical distributions and non-gaussian noise measeruments?

[ismail bsa]

Hello,
I have some non Gaussian noise measurements and I want to find a statistical model that best represents those measurements.

The statistical models that I used for comparison are Middleton Class A , Alpha stable and the normal Gaussian distributions.

First I calculated the PDF and the CCDF of the measurements and of each one the models.

Then I compared the models with the measurements by calculating the Kull-Back Libeler divergence and by calculating the MSE between the PDF of measures and those of the statistical models.

I also calculated the power spectral density (PSD) of measurements the PSD of the noise resulting from each one the models.

Do you know other methods or metrics to do such comparison?

Thanks.

 

[mrflibble]

First I calculated the PDF and the CCDF of the measurements and of each one the models.

Do you have some plots of the PDF + CCDF of your measurements? That might help in getting you some ideas for new statistical models you can use to represent your measurements. All we know now is "it's not Gaussian". And again for all we know it could be almost-but-not-quite Gaussian, which still is non-Gaussian, but in which case I would probably treat it as Gaussian anyways. So some plots would really help.