Equal Sized random assortment, using randi?

Hello everyone,

I have used the following code to generate a matrix of random integers (1's and 2's). However, I am finding myself with an unequal amount of random integers in each column (e.g., eleven "1's" and eight "2's" in some columns). I would like to know how I could get a fixed amount of equally sized conditions using this function. In where, I could get ten "1's" and ten "2's" equally spread among all six columns, that still maintains a random order. Am I using the right function (randi) to do accomplish this? Please let me know if I am unclear with my question.

Kind regards,

T = [ ];
ii=1;
while ii<=20;
    T(ii,:)= randi(2,1,8);
    ii=ii+1;
end;

Answer 1

[Image Analyst]

Chris: first you need to create a column vector with the desired number of 1s and 2s in the column. Then you need to use randperm() to scramble the order and stick it in your output array:

num1s = 10; % Whatever you want
num2s = 10; % Whatever you want
% Make a column vector with the specified number of 1s and 2s in it.
unscrambled = [ones(num1s, 1); 2 * ones(num2s, 1)]
rows = length(unscrambled);
columns = 16; % Whatever you want
output = zeros(rows, columns);
% Scramble and place into the output matrix.
for col = 1 : columns
	sortOrder = randperm(rows);
	output(:, col) = unscrambled(sortOrder);
end
output % See it in the command window.

[Chad Greene]

Ha. Great minds think alike.

Evidently ours do too.

[Chris Keightley]

Wow truly incredible from the both of you!! Thank you so much

[Chad Greene]

This line

rows = length(unscrambled);

will not work as desired if (num1s+num2s)<columns. To avoid a potential error I'd change it to

rows = size(unscrambled,1); 

[Image Analyst]

Chad, I don't understand what you're saying. If I change it to have

num1s = 4; % Whatever you want
num2s = 3; % Whatever you want
% Make a column vector with the specified number of 1s and 2s in it.
unscrambled = [ones(num1s, 1); 2 * ones(num2s, 1)]
rows = length(unscrambled);
columns = 9; % Whatever you want
output = zeros(rows, columns);
% Scramble and place into the output matrix.
for col = 1 : columns
	sortOrder = randperm(rows);
	output(:, col) = unscrambled(sortOrder);
end
output % See it in the command window.

so that 4+3 < 9, it still works just fine with the specified number of 1s and 2s in each column.

output =

2 1 2 2 1 1 1 1 1

1 1 2 1 2 1 1 1 1

1 2 1 1 1 2 1 1 2

2 1 1 1 1 1 2 2 1

1 1 1 2 2 1 1 2 1

2 2 2 1 1 2 2 2 2

1 2 1 2 2 2 2 1 2

And since unscrambled is a column vector, length(unscrambled) is the same as size(unscrambled, 1). How would there be an error?

[Chad Greene]

Oh heck, you're right @Image Analyst. I misread, thinking that unscrambled was a matrix.

Answer 2

[Chad Greene]

This would be one way to define how many ones and how many twos in each column before randomizing them:

N_cols = 8; % number of columns
N_ones = 10; % number of ones in each column
N_twos = 10; % number of twos in each column

M = [ones(N_ones,N_cols);2*ones(N_twos,N_cols)]; 

imagesc(M)

% Shuffle the order of each column: 
for k = 1:N_cols 
   M(:,k) = M(randperm(N_ones+N_twos),k); 
end

imagesc(M)

[Chris Keightley]

Wow, truly mind-blowing, thanks a lot Chad, love the pictorial demonstration as well.