Quantile–Quantile Plot
A diagnostic plot comparing quantiles of two distributions at matching cumulative probability levels.
A Q–Q plot pairs quantiles evaluated at matching cumulative probability levels in two distributions. One axis may use theoretical quantiles from a reference model and the other sample quantiles. Its purpose is to compare distributional shape, location and scale.
Interpreting a straight pattern
If two distributions differ only by location and scale, their corresponding quantiles often form an approximately straight line. Alignment with y=x is a stronger requirement, indicating matching quantiles on the same scale. A different slope may indicate a scale difference; an intercept shift may indicate different location.
In a normal Q–Q plot, the reference quantiles come from the normal distribution. Systematic curvature in the tails suggests disagreement in tail behavior. S-shaped departures can be associated with lighter or heavier tails, but exact interpretation depends on the orientation of the axes and the quantile estimator.
Diagnostic limits
Extreme sample quantiles are uncertain in small samples; random curvature should not be overinterpreted. With very large datasets even subtle discrepancies may become statistically detectable. A Q–Q plot is a diagnostic, not conclusive evidence that a distribution is or is not normal. It can be considered together with an ECDF or suitable formal tests.
A histogram groups values by intervals. A Q–Q plot instead matches probability levels directly. Skewness provides a separate numerical summary of asymmetry.