dplyr's ntile() is not a quantile binning function, and that's a problem

The Trouble with 'Ntile()'

In a post from a data science blog, the author warns that dplyr::ntile() does not perform quantile binning as its name suggests. Using a simulated pharmacometric dataset with many tied weights at the median, she shows that ntile() assigns tied values arbitrarily to adjacent bins, making it unsuitable for statistical analysis. She advises against using it where accuracy matters.