Simulate Data

First, we generate conditions for a simulation. These vary a sample size and a true correlation rho. Second, we define a function that simulates data for two normally-distributed variables for a given simulation condition, and returns an estimate of the Pearson correlation of the two variables:

conditions <- expand.grid(N=c(50,100),rho=c(0,0.25,0.5), iter=1:10)

compute_correlation <- function(N, rho, ...) {
  N<-100
  r=0
  data <- MASS::mvrnorm(n=N, mu=c(0,0), Sigma=matrix(c(1,r,r,1),nrow=2))
  cor(data[,1],data[,2])
}

Last, we call the function for every simulation condition to obtain the estimates for all simulation conditions.

results <- apply(conditions, 1, function(row) {
  do.call(compute_correlation, as.list(row))
})

Note that the first code chunk reproduces fine while second code chunk does not. This is because the simulation did not specify a random seed, so the random numbers are different every time the document is generated. This is caught by using the reproducibleR chunk.