Sample n times with replacement in Julia

Question

I'm trying to draw n -- say, 10 for simplicity -- samples from an array in Julia. Using the function wsample, ndraws below gives me what I wanted

using Distributions
population = [ 1, 10 ]
weights = [ .4, .6 ]
population_idx = wsample( population, weights, 10 ) # i.e. population indices
ndraws = population[ population_idx ]

I'm using Julia 0.2. Is there a way to do the same thing without the indices? In R for instance, we have

ndraws <- sample( population, size = 10, replace = TRUE, prob = weights )

The docs here suggests there is i.e. doing this

ndraws = wsample( population, weights, 10 )

should give me, err, exactly what I want? Note too, that in the docs, the argument name for number of draws is n but looking in the source code of sample.jl, it refers to k.

When I run your code, wsample( population, weights, 10 ) gives the answer you want (values of 1 and 10). When I run population[ population_idx ] I get ERROR: BoundsError(). Is there some documentation that says wsample returns an index? It appears to me to return a sample vector from population — James King
@user3114046 Thanks. Inspection of methods( wsample ) is what led to the confusion. The whole multiple dispatch business will take a while to sink in. I would've accepted your answer had it not been a comment. — vathymut

lindahua lindahua · Accepted Answer · 2014-05-26T22:35:40

wsample(population, weights, n) returns n samples from population (not the indexes). This is in the Distributions package.

For example,

julia> using Distributions
julia> wsample(["a", "b"], [0.3, 0.7], 10)
10-element Array{ASCIIString,1}:
 "a"
 "b"
 "b"
 "b"
 "b"
 "b"
 "b"
 "b"
 "a"
 "b"

If what you want is to draw indexes, you can use 1:k as the population, for example, wsample(1:k, weights, n).

Sample n times with replacement in Julia

1 Answers