What happens if you (incorrectly) treat individual Likert scales as though they're truly continuous, normally distributed responses?
awhug/shinyvariance
Shiny simulation of the change in variance (and mean) with each additional observation on a Likert scale
RGPL-3.0