Random weighted numbers without repeating

I'm trying to calculate 5 numbers at random. The numbers 1-35 have set probability weights that are assigned to each number. I'm wondering how, in Matlab, to compute 5 random numbers with weights WITHOUT repeats.

Answer 1

[Walter Roberson]

You can use datasample() from the Statistics toolbox.

If you happen to be using a version of the Statistics toolbox too old to have datasample() then see http://www.mathworks.com/matlabcentral/answers/27515-random-sample-without-replacement#answer_35757

Answer 2

[the cyclist]

If you have a recent version of the Statistics and Machine Learning Toolbox ...

M =  5; % Number in sample
N = 35; % Number in population

population = 1:N;

% Create some pretend weights. Put your real weights here.
w = rand(1,N);
w = w/sum(w);

x = datasample(population,M,'Replace',false,'Weights',w)

Answer 3

[the cyclist]

If you have a version of the S & ML toolbox that predates datasample, you could do use randsample. But it does not support non-replacement, so you may need to use the rejection method if you choose non-unique elements. How inefficient this is depends on your weights.

M =  5; % Number in sample
N = 35; % Number in population

population = 1:N;

% Create some pretend weights. Put your real weights here.
w = rand(1,N);
w = w/sum(w);

% Preallocate
y = zeros(1,M);

% Generate values until you have M with no repeats
while numel(unique(y)) ~= M
    y = randsample(population,M,true,w);
end