Using randsample to randomize rows

I have 36 households. I want to study combinations of households. PR represents how many households that are combined. To make other calculations

feasible, I sample the data using randsample. However, when I increase PR to 9 or more, the number of combinations become to great for me to sample like this. Can I integrate the sampling with nchoosek somehow, or is there another way?

nHousehold =36;9 or

PR=7 ; %Number of Households that are to be selected

Hcombos=nchoosek(1:nHousehold,PR); %Number of combinations

y = randsample(length(Hcombos(:,1)),2);

Hcombosnew=[];

for b=1:2

Hcombosnew=[Hcombosnew; Hcombos(y(b),:)];

end

[David Hill]

Do you have to test all combinations or just a random sampling of the combinations?

[Joel Schelander]

Just a random sampling, since there will be too many combinations

Answer 1

[David Hill]

You could run a loop however many times you want to generate the sampling and then make sure they are all different.

y=zeros(1e5,9);
for k=1:1e5
  y(k,:)=randperm(36,9);
end
y=unique(sort(y,2),'rows');