The while loop is a sometimes controversial programming construct since it is often not clear that the loop will ever end. Here’s one situation where I think it is appropriate to use. In this example I would like to generate a small dataset with a high correlation coefficient (\(r > 0.9\)) and a significant p-value for the correlation (\(p < 0.05\)). In order to do this I just keep generating random datasets until I find one that satisfies my criteria. This is certainly a non-elegant brute-force approach, but it gets the job done. I use this approach when I’m trying to generate a “true positive” plot to illustrate the concept of using a line up for graphical inference.



n <- 10

data <- data_frame(x = runif(n),
           y = runif(n))

test <- cor.test(data$x, data$y)

while(!(test$p.value < 0.05 && abs(test$estimate) > 0.9)) {
  data$x <- runif(n)
  data$y <- runif(n)
  test <- cor.test(data$x, data$y)

plot(data, xlim = c(0, 1), ylim = c(0, 1), asp = 1)