Empirical Cumulative Distribution Function
A step function giving the proportion of observed values less than or equal to each threshold.
An empirical cumulative distribution function (ECDF) is built directly from observed data. For observations x1,…,xn it is defined by:
F_n(t) = (1/n) × Σ I(x_i ≤ t)The indicator I equals 1 when its condition holds and 0 otherwise. The function reports the observed fraction at or below threshold t; no histogram bin width or starting edge is required.
What each jump represents
For 2, 2, 5, 9, we obtain F_n(1)=0, F_n(2)=0.50, F_n(5)=0.75 and F_n(9)=1. The repeated value at 2 produces a jump of 2/4. An ECDF is nondecreasing and right-continuous.
It can answer threshold questions, compare cumulative proportions across groups and expose the locations of observed quantiles. Quantiles ask the inverse question: what threshold corresponds to a chosen cumulative proportion?
Unlike a histogram, it has no bins
A statistical histogram displays counts in intervals and changes when those intervals change. The ECDF uses exact observed ordering; its steps remain visible with small samples. A KDE instead estimates a smooth density. None of these plots guarantees that the empirical sample perfectly represents its source population.
Right-censored survival observations call for estimators such as Kaplan–Meier rather than an ordinary uncensored ECDF. Missing-data and observation-weighting conventions also require explicit disclosure.