Rician Distribution will not fit to positive data?

I am trying to fit a Rician distribution to my data, z, that ranges between 2X10^-9 and 5X10^-8 via

pd = fitdist(z,'Rician');

however, the error

Error using prob.RicianDistribution>checkargs (line 197)
The parameter S must be a nonnegative finite numeric scalar.

always pops up. I get the exact same error via

phat = mle(z,'distribution','rician')

and the dfittool won't respond when I ask it to fit a Rician distribution to my data. I'm at a loss after trolling google for answers. Any suggestions?

[Bruno Pop-Stefanov]

Would you mind attaching the vector z to your question as a MAT file so I can take a look?

If that can be of any help, the error is thrown here:

    if ~(isscalar(s) && isnumeric(s) && isreal(s) && isfinite(s) && s>=0)
        error(message('stats:probdists:NonnegativeParameter','S'))
    end

at line 197 of prob.RicianDistribution.m.

By looking at the code, the class does some pre-processing on the data. I suppose it transforms the data so that it does not pass the test above.

[the cyclist]

I'd refine Bruno's comment by suggesting you provide the minimal vector that causes the problem. Sometimes the process itself of trying to identify the minimal data that triggers the error message will bring the core issue to the fore.

[Jason Wang]

I got the error with this 0.0005 0.0003 0.0008 0.0016 0.0002

[the cyclist]

This code:

z = [0.0005; 0.0003; 0.0008; 0.0016; 0.0002];
pd = fitdist(z,'Rician');

returns successfully for me in R2017a, using Mac OS X 10.10.5.

[abhishek sunkum rammurthy]

Even i am getting same error, someone please tell me if this problem is solved.

[John D'Errico]

Of course it is "solved", in the sense that it is not a problem. ;-)

z = [0.0005; 0.0003; 0.0008; 0.0016; 0.0002];
pd = fitdist(z,'Rician')
pd = 
  RicianDistribution

  Rician distribution
        s = 1.62329e-06   [0, 0.000157863]
    sigma = 0.000598335   [0.000372161, 0.000961963]

That error will only result IFF your data includes some points that are...

• not numeric
• complex
• inf or nan
• less than or equal to zero
• or you have only ONE data point

So you need to show data that actually causes a problem. As well, tell which MATLAB release you are trying to do this with. If you are using an older release, that may explain the issue. Otherwise, it is terribly difficult to track down a problem that does not happen.

[yusra Ch]

I have the same error. Every time the value of "s" parameter is very near to zero the error shows up. Can you tell me how to bypass this error?

[the cyclist]

I have the same comment as in the original question. Can you post an example of an input that gives an error? Does the following code give an error?

z = [0.0005; 0.0003; 0.0008; 0.0016; 0.0002];
pd = fitdist(z,'Rician');