I have the defined the following variables:
a <- as.character(1:10)
b <- 100
c <- 10
probs <- c(0.3, 0.3, 0.3, rep(0.1/7, 7))
min <- 5
max <- 10
I am trying to figure out how to subset the 'probs' argument in the code below to correspond with the characters that are randomly sampled (i.e., randomly sampling characters 5:10)
sample(a[min:max], size = round(b/c), replace = TRUE, prob = probs[???])
I don't think that probs[min:max] will work as it should, but I am uncertain how to find out for certain if it does. A more complicated scenario is if I want something like
a[c(1, 3, 5)]
I would then need 'probs' to correspond to characters 1, 3, and 5.
I have tried using probs[get(paste0(...))], but this is not the most direct and efficient way. It doesn't work anyway.
Any advice is appreciated.
You simply need to subset probs to be the same subset as a, e.g.
index = min:max
sample(a[index], size = round(b/c), replace = TRUE, prob = probs[index])
For the more complicated scenario, set index = c(1,3,5).
You can see that this works by doing a simulation and comparing the observed probabilities with the true probabilities:
set.seed(123)
tmp = sample(a[min:max], size = 10000, replace = TRUE, prob = probs[min:max])
table(tmp)/10000 # the observed probabilities
The observed probabilities are roughly all equal, which is consistent with the true probabilities being all equal.
If you take a look at the help files for sample, you'll see that prob does not need to sum to 1. The function will take care of normalizing the probabilities.