How to select k out of n, k being normally distributed but not exceeding n

Hello everyone,

I need to select k numbers out of n without generating doubles and k must not exceed n.

I used to do it by first getting k with normrnd and a very low variation and then finding random numbers and checking for doubles until I have my set of k. Not a very nice approach I know.

Just now I found randperm and it allows me to do this more elegantly, but the prime use of that function, the permutation, is wasted effort in my use. I also switched from normrnd(mu,sigma) to mu+sigma*randn.

Does anyone know a more fficient way of doing this?

Thank you.

Answer 1

[Matt J]

There is nchoosek(n,k).

It's not clear, however, what "k being normally distributed" is supposed to mean. Clearly, k is always a positive integer, so it cannot be normally distributed.

[Daniel]

Thank you for the first part.

k should be normally distributed integer over the range 1 to n-1. Initially I ran into problems when k exceeded n and I solved it by decreasing the standard deviation. I am not happy with this, as it still leaves the problem unsolved. Do you have an idea how to do this better?

Edit: nchoosek is sadly not what I need. I don't need the binomial coefficient. I need one set of k numbers froom 0 to n.

[Matt J]

I still have no idea what a "normally distributed integer" means. It is definitely not standard terminology. Perhaps you mean a uniformly distributed integer, in which case you could use RANDI.