ecdf

Distributions

Minimal call

import random
rng = random.Random(8)
df1 = {"x": [rng.gauss(0, 1) for _ in range(200)]}
df2 = {"x": [rng.gauss(0.6, 1.3) for _ in range(200)]}

c = pt.chart(xlabel="value", ylabel="F(x)")
c.add_ecdf(df1, aes(x="x"))
c.add_ecdf(df2, aes(x="x"))

Worked example

import plotlet as pt
from plotlet import aes

df = pt.load_dataset("tips")

c = pt.chart(df, aes(x="total_bill", color="time"),
             title="bill size by service", xlabel="total bill ($)",
             ylabel="fraction of parties",
             data_width=320, data_height=210)
c.add_ecdf()

Docstring

Empirical CDF as a step function — no bin choice, every observation visible.

F̂(x) = (#{xi ≤ x}) / n as a step function. ECDFs are the statistician-
preferred alternative to histograms: no bin choice, no smoothing, every
observation visible — overlaying multiple groups makes distribution
differences obvious.

  c.add_ecdf(aes(x="col"))                      # columns via aes
  c.add_ecdf(aes(x="col", color="group"))       # one curve per group

Aesthetics:
  color=         bare → literal line color; aes(color="col") → grouped curves
  palette=       maps group levels → colors when color is mapped in aes

Other styling kwargs:
  complement=False   True draws 1 - F̂(x) (survival function)
  linewidth=1.5      stroke width
  label=None         legend label (single-series only)
  simplify=None      None = auto-drop step vertices the stroke can't show
                     above `dense_threshold` observations (the curve has
                     ~2 vertices per observation); True/False forces it
                     on/off (see draw/_simplify.py)