This function calculates the number of permutations of a multiset, this being the multinomial coefficient. If a set \(X\) contains \(k\) unique elements \(x_1, x_2, \ldots, x_k\) with associate counts (or multiplicities) of \(n_1, n_2, \ldots, n_k\), then this function returns $$\frac{n!}{n_1!n_2!\ldots n_k!}$$ where \(n = \sum_{i=1}{k}n_i\).

`multinom(x, counts = FALSE, useDouble = FALSE)`

x

Either a multiset (with one or more potentially non-unique
elements), or if `counts`

is `TRUE`

a set of counts of the unique
elements of \(X\). If `counts`

is `FALSE`

and `x`

is not
numeric, then x will be coerced into an integer vector internally. If
`counts`

is `TRUE`

then `x`

must be a vector of integers that
are greater than, or equal to zero.

counts

if `counts`

is TRUE, then this means x is the set of
counts \(n_1, n_2, \ldots, n_k\) rather than the set itself

useDouble

if `useDouble`

is TRUE then the computation will be
done using double precision floating point arithmetic. This option was added
because the internal code cannot handle integer overflow. The double
precision code will may a result that is closer to the truth for large
values, but this is not guaranteed. Ideally something like the GMP library
should be used, but this is not a priority at this point in time.

A single integer representing the multinomial coefficient for the given multiset, or given set of multiplicities.

multinom depends on C++ code written by Dave Barber which can be found at http://tamivox.org/dave/multinomial/code.html. The code may require the STL algorithm library to be included in order to compile it.

# NOT RUN { ## An example with a multiset X = (a,a,a,b,b,c) ## There are 3 a s, 2 b s and 1 c, so the answer should be ## (3+2+1)!/(3!2!1!) = 6!/3!2!1! = 60 x = rep(letters[1:3],3:1) multinom(x) ## in this example x is a vector of counts ## the answer should be the same as above as x = c(3,2,1) x = rep(letters[1:3],3:1) x = as.vector(table(x)) #coerce x into a vector of counts multinom(x, counts = TRUE) ## An example of integer overflow. x is a vector of counts ## c(12,11,8,8,6,5). The true answer from Maple is ## 11,324,718,121,789,252,764,532,876,767,840,000 ## The error in the integer based answer is obvious. ## The error using floating point is not, but from Maple is ## 0.705057123232160000e+10 ## Thanks to Lev Dashevskiy for calling my attention to this. # } # NOT RUN { x = c(12,11,8,8,6,5) multinom(x, counts = TRUE, useDouble = FALSE) multinom(x, counts = TRUE, useDouble = TRUE) # } # NOT RUN { # }