I'd like to create a normal probability plot like the one below and was wondering if it has a particular name? I ask so that I can search for that name and find help on how to create a graph just like it.

I got this graph from a book I'm reading and the author says:

When more than 10% of the dots fall outside the blue, lines, there is reason to suspect the data is not normal.

This easy measure of non-normality is the reason I I'm interesting in finding out how to create this sort of normal probability plot - it seems easier than other approaches for readers to interpret because it provides something to quantify (as opposed to visual judgement of whether the data just looks normal).

enter image description here

  • $\begingroup$ This seems rather arbitrary. Why is 9.9% OK? If you think normality testing is essentially useful, & want something to quantify, why not use a standard test for normality like the Shapiro-Wilk? $\endgroup$ Sep 8, 2016 at 19:06

1 Answer 1


This is a Q-Q plot with confidence intervals. The plot you posted was created with Minitab, but you could create the same in R.


  • $\begingroup$ qq-plots compare quantiles. The Y axis is labeled "percent". I can't tell if this is a standard qq-plot with a different axis pasted on, or a non-standard plot. $\endgroup$ Sep 8, 2016 at 21:03
  • $\begingroup$ @gung it's a Q-Q plot with the y-values (approximately) as the expected quantiles for sampling from a standard normal $\Phi^{-1}(p)$ but labelled with the percentile ranks ($p$). That labelling makes it look like the old normal-probability-paper that sufficiently ancient statisticians may recall (from when such plots were made by hand rather than computer). The y-axis labels bothers me less than having the random variable on the x-axis; that makes me twitch. $\endgroup$
    – Glen_b
    Sep 9, 2016 at 0:16

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.