A normal probability plot is extremely useful for checking normality assumptions. It’s more precise than a histogram, which can’t pick up subtle deviations. And yet it doesn’t suffer from too much power from large samples with tiny departures from normality or too little power from small samples with large departures from normality, as do tests like Shaprio-Wilkes.
The biggest problem with a normal probability plot is that it’s hard to read, especially if you’re not used to them. So let’s take a moment and walk through exactly how they work and what they tell you.
There are two versions of normal probability plot: Q-Q and P-P. I’ll start with the Q-Q. (more…)