creating weighted sample for a matrix

I need to create a sample from an m by n matrix. Each row has a probability associated with it and can be treated as one observation. I need to draw N (the sample size) number of these rows based on their assigned probability. The problem is that randsample only takes a vector as input such as this example:

R = randsample([1 2 3], N, true, [0.5 0.25 0.25])

Whereas I need something like this:

    R = randsample([1 2 3; 4 5 6; 7 8 9], N, true, [0.5 0.25 0.25])

So far I've not found an answer from Google, so I'd appreciate any pointers.

Answer 1

[Sophia]

Thanks Sean, I figured it out. I add a unique ID to the first column of the sample matrix (route_sample) and perform randsample on that; then I use the FIND function to pull the actual sample values in.

   %create random exogenous demand sample
   R_tmp = randsample(route_sample(:,1),N,true,route_sample(:,8));
   R_demand = zeros(N,size(route_sample,2)-2);

   for i = 1:size(R_tmp,1);
   [r1,~,~] = find(R_tmp(i,1)==route_sample(:,1));
   R_demand(i,:) = route_sample(r1,2:7);
   end;

[Michele Pio Papasidero]

Dear Sophia, i have a similar issue (https://it.mathworks.com/matlabcentral/answers/812345-weighted-random-sampling-for-matrix). I tried to adapt this code to my situation but it does not work. Can you help me?

Answer 2

[Sean de Wolski]

(Unless I'm missing something (which is certainly possible (probable))) This could easily be done with a for-loop.