MerQuR+Statistics Glossary

Statistics Glossary → Q-Q and P-P plots

Q-Q and P-P plotsconcept

Plots that show how well data follow a theoretical distribution (usually the normal) by comparing observed and expected quantiles (Q-Q) or cumulative probabilities (P-P).

In a Q-Q (quantile-quantile) plot the ordered observations are plotted against the values expected at the same ranks under a normal distribution. The expected values use a plotting-position formula such as Blom's (i − 3/8)/(n + 1/4). If the data are normal, the points lie roughly on a straight line. Points that bend away from the line in opposite directions at both ends indicate heavy tails, a curve bending one way indicates skewness, and isolated distant points indicate outliers. A detrended Q-Q plot magnifies these patterns by plotting the deviations from the line around a horizontal axis.

A P-P plot compares the cumulative probabilities of the observations with the theoretical cumulative probabilities; it is more sensitive to departures in the middle of the distribution and less sensitive in the tails. Q-Q plots are generally preferred because extreme values and tail behaviour matter more for most methods. These plots are not tests: in large samples normality tests flag even trivial departures, whereas the plot shows how large the departure is and what form it takes. In regression and ANOVA the plot should be applied to the residuals, not to the raw data.

An everyday example

A café examines the preparation time of 200 orders. In the Q-Q plot most points lie on the line, but a few at the right end rise clearly above it. This shows that a few orders took much longer than expected, so the distribution is right-skewed with a long tail.

In MerQur

Descriptive statistics offers 'Normal curve on histogram and Q-Q' as chart extras, and Normality Tests has a 'Second chart' option (Q-Q, Detrended Q-Q, P-P, Box plot). The 'Chart' option draws a Normal Q-Q plot in the independent-samples t-test and in one-way ANOVA (per group), a Q-Q plot of the differences in the paired t-test and a Q-Q plot of residuals in two-way ANOVA. 'Additional output: Normality of differences' in Bland-Altman analysis and the 'Diagnostic plot' in multiple linear regression also include a Q-Q plot.

Analyses where this concept appears

Descriptive StatisticsNormality TestsIndependent-Samples t-TestPaired t-TestOne-Way ANOVATwo-Way ANOVABland-Altman Agreement AnalysisMultiple Linear Regression

Related concepts

Normality assumption