I'm confronted with a probability problem that may have already been solved.

I'm considering random sequences of m events. The number of possible events is n, and each type of event has the same fixed probability of occurrence 1/n. I would like to know the probability distribution of the number of different events in the sequence (comprised between 1 and min(n,m))


Indeed, the problem was already solved. It's a generalization of Birthday paradox. The formula for probability mass function is

The probability of getting $k$ unique values from $[0, n)$ when choosing $m$ times is given by:

$$P(V = k) = \binom{n}{k}\displaystyle\sum_{i=0}^k (-1)^i \binom{k}{i} \left(\frac{k-i}{n}\right)^m $$

where $V$ is a random variable giving the number of unique outcomes and $\binom{\cdot}{\cdot}$ is the binomial coefficient.

as quoted by George V. Williams on math.stackexchange.com who refers to http://www.math.uah.edu/stat/urn/Birthday.html#General


m <- 100
n <- 40

f <- function(x, m, n) {
  vapply(x, function(k) {
    choose(n, k) * sum( ((-1)^(0:k) * choose(k, (0:k)) * ((k - (0:k))/n)^m) )
  }, numeric(1))

out <- replicate(1e6, {
  length(unique(sample.int(n, m, replace = TRUE))) 

plot(prop.table(table(out)), xlab = "k", ylab = "P(V=k)")
lines(1:m, f(1:m, m, n), col = "red")

Simulated values vs true pmf

| cite | improve this answer | |
  • $\begingroup$ +1 Is it possible that the use of $n$ and $m$ is swapped in the OP with respect to the answer? $\endgroup$ – Antoni Parellada Oct 31 '16 at 10:54

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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