How can i do probability distribution fitting of experimental data?

Hi Everybody, I would like to do Fit with 'Rayleigh', 'Weibull', 'Lognormal', 'Nakagami','Rician' of experimental data those are in negative log values;

Problem: Once i do fitting like pd2 = fitdist(vecteur_amplitude','Rician'); gives

Error using prob.RicianDistribution>ricefit (line 273) The data in X must be positive

This error for all distributions except 'Rayleigh'. Could you please help me. Thanks.

Regards Tariq

Answer 1

[the cyclist]

The error message seems pretty straightforward. A Rician distribution, by definition, has support only over positive values of x. If your data are negative, then they cannot possibly be directly fit with a Rician. Did you instead want to fit a Rician to -x?

Answer 2

[Shah Tariq]

Hi, Thanks for your comments. Yes, Actually i would like to do the fitting as in the attached article plotted in the figure 7 or 8. Please find the attached file. He fitted negative values with all distributions. Once i use fitdist function, Except 'Rayleigh', other distributions 'Weibull', 'Lognormal', 'Nakagami','Rician' have the same error. If it does not work, how he did. Moreover, how can i do the mean to zero for fitting.

Please help me. Thanks in advance.

Regards Tariq

[the cyclist]

I took a quick look at the paper. I don't believe they fit any negative values. As they describe in the paper, they did a transformation on the data to display it. I expect that they started off with entirely positive values which were fit. Then they normalized to the median value (presumably meaning that they simply divided by the median), and reported that ratio in decibels, which will mean that small positive values of the ratio will have negative values in decibels.

But, that's just my best guess as to what they did.