I am interested in computing the probability of selecting a specific sample say $u_{1}, u_{2},...,u_{n}$ from a population having $N$ sampling units. I know that in without replacement there are total number of possible sample as $N \choose n$ and hence the probability of such selection will be inverse of this.

Similarly, in the with replacement case, the no of such samples will be $N^{n}$ and consequently the probability of such selection will be inverse of this.

But if I try to view this problem as computing the $P(u_{1}, u_{2},...,u_{n})$. I can write:

$P(u_{1}, u_{2},...,u_{n}) = P(u_{1})P(u_{2})....P(u_{n}) $. Let us call it Equation (1)

In the case of with replacement, $P(u_{i}) = \frac{1}{N}$. Putting this in equation (1), It gives us the same result as the previous method gave. But in the situation of Without replacement, we have:

$P(u_{i}) = \frac{1}{N - (i-1)}$.

It is not giving the result which is consistent with the previous one.

Can someone suggest me what is it that I am doing wrong here?


1 Answer 1


In the first way, you are not considering the order, but in the second way, you are considering the order that they are chosen.
$N\choose n$ is the number of ways to pick a sample of size $n$ from the population without replacement. But, then there are $n!$ different ways to arrange those objects. So, the probability of selecting them in that order is $$\frac{1}{{N\choose n}n!}=\frac{1}{N (N-1)...(N-(n-1))}$$


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.