MATLAB Answers

Alex
0

How to create multidimensional arrays of distribution objects?

Asked by Alex
on 2 Oct 2013
Latest activity Commented on by Alex
on 3 Oct 2013

Hi I am just working on a larger program that works with two random distributed Variables:

X for incoming products(Production) Y for outgoing products(Demand)

both random variables have 2 parameters, so a and b belong to X and c and d belong to Y. I am now working on a combination of both and need a combined distribution stored in a 4 dimensional array to reflect all possible combinations. My initial guess to create an "empty" array of the matching dimensions was just:

    a=[1 2 3 4];
    b=[5 6];
    c=[7 8 9];
    d=[1 2 3 4 5];
    for i=1:length(a)    
        for j=1:length(b)    
            for k=1:length(c)    
               for l=1:length(d)    
                   test(i,j,k,l)=makedist('normal',0,0);    
               end
            end
        end
    end

the problem is now that matlab seems to have no default distribution-object included in the libraries and following error occurs:

During array expansion: No default is defined for class 'prob.NormalDistribution'. Method 'getDefaultScalarElement' in superclass prob.ProbabilityDistribution is missing or incorrectly defined.

I was lucky find a solution for 2 dimensions that just sets the elements before a default element is needed and i expanded this to a 3 dimensional array. However, if i try to expand the solution in the same way to 4 dimensions the error occurs again:

clc  
clear  
a=[100 5 5 7];   
b=[10 4 4 5 6];   
c=[100 2 6 7 9 9];   
d=[1 4 5 5]; 
for i=1:length(a)   
    test(i,:)=makedist('normal',0,0);  
    for j=1:length(b)  
        test (i,j)=makedist('normal',0,0);  
    end  
end
test
for i=1:length(a)   
    test2(i,:,:)=makedist('normal',0,0);  
    for j=1:length(b)  
        test2 (i,j,:)=makedist('normal',0,0);  
        for k=1:length(c)   
            test2 (i,j,k)=makedist('normal',0,0);             
        end  
    end  
end
test2
for i=1:length(a)   
    test2(i,:,:,:)=makedist('normal',0,0);  
    for j=1:length(b)  
        test2 (i,j,:,:)=makedist('normal',0,0);  
        for k=1:length(c)   
            test2 (i,j,k,:)=makedist('normal',0,0);   
            for l=1:length(d)  
                test2(i,j,k,l)=makedist('normal',0,0);  
            end  
        end  
    end  
end
test3

While the first two versions are running perfectly, even as intended and give the matching distribution-object arrays, but the 4 dimensional one gives following error:

During array expansion:  
No default is defined for class 'prob.NormalDistribution'.  
Method 'getDefaultScalarElement' in superclass prob.ProbabilityDistribution is missing or incorrectly defined.
Error in Untitled (line 29)
    test3(i,:,:,:)=makedist('normal',0,0);

When i define the 3 dimensional array first, set test3=test2, the error shifts to the line

test3(i,j,k,l)=makedist('normal',0,0);

as soon as d=[ ] has more than one element. i also tried to set the d loop around the other 3 loops but the error always occurs when the pointer of the fourth dimension shifts to 2. I dont know why this error occurs only with 4 dimensions, to my understanding it should also occur with 3 dimensions, or both versions should run propperly, i cant see why one works and the other one not.

It would be very kind if somebody could help me with this problem or give me a hint how to create this AxBxCxD array of distribution objects in another way. Thank you for your help

  2 Comments

Matt J
on 2 Oct 2013

Please use the

toolbar button to format your code (and error messages) more readably.

Alex
on 2 Oct 2013

Sorry, I didnt see that button, i reformated it, thank you

1 Answer

Answer by Tom Lane
on 2 Oct 2013
 Accepted answer

This looks like something that we should fix in a future release. Meanwhile, perhaps you would find this work-around useful:

a = makedist('normal');           % create a single distribution object
a = repmat(a,[2 3 4 5]);          % expand it to the desired size
a(2,3,4,5) = makedist('Weibull'); % fill in the values you really want

  1 Comment

Alex
on 3 Oct 2013

Ah great, that is the perfect solution, thank you very much!


Discover MakerZone

MATLAB and Simulink resources for Arduino, LEGO, and Raspberry Pi test

Learn more

Discover what MATLAB® can do for your career.

Opportunities for recent engineering grads.

Apply Today

MATLAB Academy

New to MATLAB?

Learn MATLAB today!