Reputation: 124
I'm trying to create a vector whose elements add up to a specific number. For example, let's say I want to create a vector with 4 elements, and they must add up to 20, so its elements could be 6, 6, 4, 4 or 2, 5, 7, 6, whatever. I tried to run some lines using sample()
and seq()
but I cannot do it.
Any help appreciated.
Upvotes: 2
Views: 68
Reputation: 32548
foo = function(n, sum1){
#Divide sum1 into 'n' parts
x = rep(sum1/n, n)
#For each x, sample a value from 1 to that value minus one
f = sapply(x, function(a) sample(1:(a-1), 1))
#Add and subtract f from 'x' so that sum(x) does not change
x = x + sample(f)
x = x - sample(f)
x = floor(x)
x[n] = x[n] - (sum(x) - sum1)
return(x)
}
Upvotes: 2
Reputation: 94182
To divide into 4 parts, you need three breakpoints from the 19 possible breaks between 20 numbers. Then your partitions are just the sizes of the intervals between 0, your partitions, and 20:
> sort(sample(19,3))
[1] 5 7 12
> diff(c(0, 5,7,12,20))
[1] 5 2 5 8
Test, lets create a big matrix of them. Each column is an instance:
> trials = sapply(1:1000, function(X){diff(c(0,sort(sample(19,3)),20))})
> trials[,1:6]
[,1] [,2] [,3] [,4] [,5] [,6]
[1,] 3 1 8 13 3 2
[2,] 4 7 10 2 9 5
[3,] 2 11 1 4 3 7
[4,] 11 1 1 1 5 6
Do they all add to 20?
> all(apply(trials,2,sum)==20)
[1] TRUE
Are there any weird cases?
> range(trials)
[1] 1 17
No, there are no zeroes and nothing bigger than 17, which will be a (1,1,1,17) case. You can't have an 18 without a zero.
Upvotes: 2