Quantile
A distribution threshold associated with a specified cumulative probability level.
A quantile describes a distribution through a cumulative probability level. For a cumulative distribution function F(x), one widely used mathematical definition is, for 0<p<1:
Q(p) = inf{x : F(x) ≥ p}The generalized inverse handles discrete distributions where an equation of the form F(x)=p might not have a solution. Q(0.5) is the median, Q(0.25) the lower quartile and Q(0.75) the upper quartile. The 90th percentile corresponds to p=0.90.
Sample quantiles need a convention
Take the sorted observations 2, 4, 6, 8. A method choosing an order statistic may give 2 at p=0.25. Another common method linearly interpolates at position h=1+(n−1)p; here h=1.75, so the result is 2+0.75×(4−2)=3.5. Different answers reflect different estimators, not necessarily defective software.
Quantiles in analysis
The interquartile range is Q(0.75)−Q(0.25). An ECDF displays cumulative observed proportions. A Q–Q plot pairs quantiles at corresponding probability levels, while quantile binning uses estimated quantiles as interval boundaries.
A reproducible report should identify the percentile estimator, missing-value policy and any observation weights. These choices matter particularly with small samples and repeated values.