What Is a Distribution, Actually?

"A distribution is a recipe for surprise. It says: when you draw one value, how surprised should you be by each possible answer?"

The Word Means More Than You Think

When a statistics textbook says 'the distribution of X,' it doesn't mean a list of values. It means the complete rulebook describing how often each value occurs (for discrete X) or how densely values pack together (for continuous X). A distribution is the full probabilistic biography of the variable. Mean and variance are summaries; the distribution itself is the whole story.

Two Flavors

Discrete: the variable takes countable values — number of heads in 10 coin flips, number of customers walking into a store in an hour. Each possible value has a probability. The collection is called a probability mass function (PMF). Probabilities sum to 1.

Continuous: the variable takes values from a continuum — a person's height, a stock's price tomorrow. Individual values have probability zero (there are infinitely many heights). What has probability is a range of values. The function that describes the density is the probability density function (PDF). The area under the PDF integrates to 1.

Discrete or continuous, the rule is the same: distributions describe how probability is spread across the range of possibilities.

The Histogram Is the Empirical Distribution

The histogram you see in everyday data plots is a finite-sample approximation of the underlying distribution. You bin the observed values, count how many fall in each bin, and the resulting bar heights estimate the true PMF/PDF. With enough samples, the histogram converges to the real distribution. With few samples, it lies.

That convergence is itself the Law of Large Numbers, which the next track will use to derive the bell curve. The distribution-from-histogram move is the citizen's first contact with the abstract idea — and most people stop there. The full quest is what you can do once you have the distribution itself, not just the bars.

Why You Cannot Skip the Word 'Distribution'

A summary statistic (mean, median, variance) without the underlying distribution is a postcard, not a photograph. The same mean can come from a normal, a power-law, a bimodal, a long-tailed distribution — and the right decision changes completely depending on which shape produced the mean. The citizen who acts only on summary statistics is acting on postcards.

Where This Goes Next

The next three lessons introduce the three citizen-relevant shapes: normal (lesson 3), skewed (lesson 4), and power-law (lesson 5). The last lesson of this track (6) names the fat-tailed family as a preview for Track 07. Then Track 03 explains why the normal shows up so often — and Track 07 explains where assuming it will silently destroy you.

Code

Three different shapes, all with the same mean·python

import numpy as np
rng = np.random.default_rng(5)

# Three samples from three very different distributions, all with mean ~ 100.
N = 10_000
normal_like   = rng.normal(loc=100, scale=15, size=N)
bimodal       = np.concatenate([rng.normal(70, 5, N // 2), rng.normal(130, 5, N // 2)])
power_law     = (rng.pareto(a=1.5, size=N) + 1) * 30   # heavy-tailed, mean ~100

for name, x in [("normal-like", normal_like), ("bimodal", bimodal), ("power-law", power_law)]:
    print(f"{name:>12s}: mean={x.mean():6.1f}  median={np.median(x):6.1f}  "
          f"std={x.std():6.1f}  max={x.max():7.1f}")

# All three have a mean near 100. The mean alone tells you nothing about
# which shape produced it. The median, the std, and the max start to give
# the shape away. A picture of the histogram tells you the whole story.
# The point: the summary is a postcard. The distribution is the photograph.

Exercise

Take three numbers from your own life — your monthly food spend, your monthly subway rides, your monthly hours of deep work. Sketch by hand what you think each distribution looks like (rough histogram). For each, write one sentence: would the mean of one month represent a 'typical' month, or would the median be more honest? Try to feel the difference between shapes before formulas show up.

Hint

Counts of similar small events (subway rides) tend to be roughly normal-ish around your routine. Spending tends to be right-skewed (occasional big nights out). Deep-work hours often look bimodal (good weeks and bad weeks).

The Word Means More Than You Think

Two Flavors

The Histogram Is the Empirical Distribution

Why You Cannot Skip the Word 'Distribution'

Where This Goes Next

Code

External links

Exercise

Progress

Comments 0