generate a random matrix with constraints

hello,

I break a question about a problem I have to solve but I do not know if what I want to do is possible. I have a matrix (193.1) I will call A, is a column vector. I want to generate a matrix (193.3) or the sum of line 1 is equal to the value of A (1,1). the values taken by A are between 0.5 and -0.5, it's continues values. My matrix A does not always have the same value or complexity of the problem. If someone has an idea that would help me a lot. Thank you.

Answer 1

[Walter Roberson]

http://www.mathworks.com/matlabcentral/fileexchange/9700-random-vectors-with-fixed-sum

[vachelard]

sory Walter, I just realized that the links led me to a function. I try to run the function but I do not. What should I do to make it work?

[Walter Roberson]

Use the download link to download a .zip. Unzip it into a convenient directory (but not under the MATLAB installation directory.) Add the path of the directory that results into the MATLAB path using pathtool. After that in your code,

cols_to_generate = 3;

for K = 1 : length(A)
   newA(K,:) = randfixedsum(1, cols_to_generate, A(K), -0.5, 0.5);
end

and then your newA will be 193 x 3 of entries between -0.5 and +0.5, with the sum of the K'th row of newA being A(K)

[vachelard]

thank you very much for you help!!

[vachelard]

when i use the function i have a message error that says:

"Error in randfixedsum (line 95) x = (b-a)*x(p+repmat([0:n:n*(m-1)],n,1))+a; % Permute & rescale x"

I do not know how to solve the problem

[Image Analyst]

Does it give you the error message? Why not tell us? Otherwise, step into the code and at that line, see what the values of all the variables are. So, here is how to solve your problem: http://blogs.mathworks.com/videos/2012/07/03/debugging-in-matlab/

[vachelard]

I'm really sorry I started to make use of matlab and I really difficulties. I do not see how I can change the function i see the line where is my error but i don't know how to solve even with your tuto

[Image Analyst]

Hang in there. Don't give up on MATLAB yet. It's not that different than debugging or using any other programming language. You'll learn eventually. Now, did you click in the margin of line 95 to set a breakpoint? A red spot will appear. Run your program and it will stop there with an arrow pointing at that line. Then hover the cursor over the variables in your code editor window, or look in the workspace panel, or type their name into the command window, to see what values they have. What are they?

[vachelard]

My error line is as follows:

x = (b-a) * x (p + repmat ([0: n: n * (m-1)] n, 1)) + a; % Switches & rescale x

Where x is [0.5152 0.5152 0.5152]

a = -0.5

b = 0.5

n = 1

m = 3

[vachelard]

sory and p=[3 1 2]

[vachelard]

Image analyst can i show you something? please

[Image Analyst]

You are missing a comma after (m-1)] and before the n. Try it like this:

index = p + repmat ([0: n: n * (m-1)], n, 1)
x = (b-a) * x (index) + a; % Switches & rescale x

Now that will also bomb if index is some fractional number, a negative number, or zero, so be sure you know what indexes you ever might expect to get.

[vachelard]

just to be sure I get the correct code. Rhedgefund = Y (:, 1);

cols_to_generate = 3;

for K = 1: length (Y)

    Neva (K, :) = randfixedsum (1 cols_to_generate, Y (K), -0.15, 0.15);

end

or Y is my data matrix 193 * 1 which is my target matrix.

Answer 2

[vachelard]

Your two answers are good but not exactly give me the desired result. I'll try to explain better.

I expected returns monthly for 15 years. I try to generated a matrix whose sum line by line give me the monthly return of my base matrix. Knowing the values of my matrix generated can be negative. I tried this code but the values are just positive: mat = rand(193, 3); rowsum = sum(mat,2); mat = bsxfun(@rdivide, mat, rowsum);