View source: R/nimbleFunction_Rexecution.R
rankSample | R Documentation |
Takes a set of non-negative weights
(do not need to sum to 1) and
returns a sample with size
elements of the integers 1:length(weights)
, where the probability of being sampled is proportional
to the value of weights
. An important note is that the output vector
will be sorted in ascending order. Also, right now it works slightly odd syntax (see example below). Later releases of NIMBLE will contain more natural syntax.
rankSample(weights, size, output, silent = FALSE)
weights |
A vector of numeric weights. Does not need to sum to 1, but must be non-negative |
size |
Size of sample |
output |
An R object into which the values will be placed. See example below for proper use |
silent |
Logical indicating whether to suppress logging information |
If invalid weights provided (i.e. negative weights or weights sum to 1), sets output = rep(1, size) and prints warning.
rankSample
can be used inside nimble functions.
rankSample
first samples from the joint distribution size
uniform(0,1) distributions by conditionally sampling from the rank statistics. This leads to
a sorted sample of uniform(0,1)'s. Then, a cdf vector is constructed from weights. Because the sample of uniforms is sorted, rankSample
walks
down the cdf in linear time and fills out the sample.
Clifford Anderson-Bergman
set.seed(1)
sampInts = NA #sampled integers will be placed in sampInts
rankSample(weights = c(1, 1, 2), size = 10, sampInts)
sampInts
# [1] 1 1 2 2 2 2 2 3 3 3
rankSample(weights = c(1, 1, 2), size = 10000, sampInts)
table(sampInts)
#sampInts
# 1 2 3
#2434 2492 5074
#Used in a nimbleFunction
sampGen <- nimbleFunction(setup = function(){
x = 1:2
},
run = function(weights = double(1), k = integer() ){
rankSample(weights, k, x)
returnType(integer(1))
return(x)
})
rSamp <- sampGen()
rSamp$run(1:4, 5)
#[1] 3 3 4 4 4
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.