Setting min/max limits to the function random (not rand nor randn)?

I am using the function RANDOM to generate random numbers with a specific Kernel PDF that I created using FITDIST. Is there a way to control the min/max of the generated random numbers?

pd1 = fitdist((data(:,1)),'Kernel');

Y1 = random(pd1,100000,1);

Answer 1

[jgg]

The easiest way to do this in your context is to restrict the actual PDF you're using to generate the random variable, rather than do a resampling of the random number. You can do this by calling the Support property:

 pd = fitdist(x,'Kernel','Support',[100,250])

Where the bound of the support are in this example [100,250]. The random function will then not generate numbers outside this range.

[Karim Shalash]

Dear jgg,

Thank you for your time. I am not sure what is the "support" property I could not find anything straightforward on help. However, I have tried the line you wrote and I get the following error:

Error using prob.KernelDistribution (line 126) SUPPORT must be 'unbounded', 'positive', or a finite two-element sorted vector.

Error in prob.KernelDistribution.fit (line 78) pd = prob.KernelDistribution(varargin{:});

Error in fitdist>localfit (line 231) pd = feval(fitter,x,'cens',c,'freq',f,varargin{:});

Error in fitdist (line 178) pd = localfit(dist,fitter,x,cens,freq,args{:});

Error in IOARDANOFF_V2 (line 148) pd1 = fitdist((stat(:,1)),'Kernel','Support',[4,3])

[jgg]

Are you sure you executed this properly? You can read the documentation on the "support" property here. Scroll down and look for "support" under the name-value pair arguments section.

For instance, this code works as expected on my PC:

 clear
 load hospital
 x = hospital.Weight;
 pd = fitdist(x,'Kernel','Support',[100,250])
 Y1 = random(pd,10000,1);

[jgg]

Ah, the issue with your code is that you should call it like:

 pd1 = fitdist((stat(:,1)),'Kernel','Support',[3,4])

The support needs to be sorted: [lb,ub]