Chapter VIII · 28 min

Discrete Families

A field guide to named stories

“To conjecture about some affair is to measure its probability.”
— Jacob Bernoulli, Ars Conjectandi (1713)

A few discrete families appear so often that they have names. Each one is a counting story that hardened into a formula. If you remember the story, you can rebuild the formula; if you remember only the formula, you will apply it to the wrong story.

Before the names, four quiet words. A distribution is the whole assignment of chance to values. A pmf (probability mass function) is that assignment written as a list $p_{X} (k) = P (X = k)$ . A family is a shape with a knob or two — $n$ and $p$ for binomial, $λ$ for Poisson. A parameter is one of those knobs. The family is the story; the parameter is how the story is tuned.

A field guide

Keep this table nearby. The question is never “which formula is famous?” It is “which story did the experiment actually tell?”

Family	Story	Lives on	Mean
Bernoulli	one yes / no	{0, 1}	p
Binomial	n independent yes/no, same p	0, 1, …, n	np
Geometric	wait for the first yes	1, 2, 3, …	1/p
Poisson	rare events, rate λ	0, 1, 2, …	λ (also the variance)
Hypergeometric	draws without replacement	0 … min(K, n)	nK/N

Bernoulli and binomial

A Bernoulli random variable is a single yes/no: $P (X = 1) = p$ , $P (X = 0) = 1 - p$ . Then $E [X] = p$ , $Var (X) = p (1 - p)$ . It is an indicator in costume.

A binomial $X \sim Bin (n, p)$ is the number of yeses in $n$ independent Bernoulli trials with the same $p$ . The counting is the classical one: there are $(k n)$ sequences with $k$ yeses, each with probability $p^{k} (1 - p)^{n - k}$ , so

P (X = k) = (k n) p^{k} (1 - p)^{n - k}, k = 0, \dots, n .

Linearity gives $E [X] = n p$ without touching the pmf. Independence of the trials gives $Var (X) = n p (1 - p)$ . If the trials are not independent, you still have a count, but you no longer have a binomial.

The binomial pmf for n = 8, p = 1/2. The shape is the row of Pascal’s triangle, normalised. Change p and the mass slides toward 0 or n.

The binomial workshop is the same picture with sliders.

Geometric waiting times

Flip until the first yes. Let $X$ be the trial number of that first yes (supported on ${1, 2, 3, \dots}$ ). Then

P (X = k) = (1 - p)^{k - 1} p, E [X] = \frac{1}{p} .

The geometric law is memoryless: given that you have already failed $m$ times, the remaining wait has the original distribution. A coin does not get embarrassed. (Some texts put the number of failures, supported on ${0, 1, 2, \dots}$ , under the same name. Check the support before you copy a mean.)

A geometric pmf with p = 0.3. Most mass sits on the early integers; the tail is a persistent (1−p) factor.

Poisson rare events

The Poisson law with rate $λ > 0$ is

P (X = k) = e^{- λ} \frac{λ ^{k}}{k !}, k = 0, 1, 2, \dots

with mean and variance both equal to $λ$ . It is the limit of $Bin (n, p_{n})$ when $n \to \infty$ and $n p_{n} \to λ$ : many trials, each unlikely, a moderate expected count. Typographical errors, arrivals at a switchboard, mutations on a strand of DNA — the same counting limit.

Poisson with λ = 3. A unimodal mass on the non-negative integers, centred near its mean, with a right tail.

Without replacement: hypergeometric

An urn holds $N$ balls, $K$ of them gold. You draw $n$ without putting them back. Let $X$ be the number of gold balls in the hand. Then

P (X = k) = \frac{( k K ) ( n - k N - K )}{( n N )},

whenever the binomial coefficients make sense, and $E [X] = n K / N$ by linearity (each position in the hand is equally likely to be gold). This is the binomial’s honest cousin: same “how many gold?”, different story, because each draw changes the urn. If $N$ is huge and $n$ is small, the two families hug — which is why people sometimes get away with a binomial on a deck of cards, and why they should not on a deck of ten.

Foundations studio: make the idea yours

This extended studio deliberately slows the pace. It is for a first-time learner who wants to recognize the idea in a new story, not merely reproduce a formula. Work with pencil and paper. Predict before calculating; redraw the pictures; and finish every numerical answer with a sentence in ordinary language.

A mental map before more algebra

A distribution family is a reusable story plus a formula. The important skill is recognizing the mechanism: one trial, fixed number of trials, wait-until-success, or count-in-an-interval.

A working map for Discrete Families. Cover the labels and reconstruct the chain from memory.

Do not treat the arrows as a theorem. They are a study aid. A strong probability habit is to move back one box whenever a formula feels unmotivated: ask what the experiment is, what information is available, and what quantity the question actually requests.

Three formulas worth being able to narrate

X s im ma t h r m B in (n, p)

Read this line from left to right and explain what every symbol refers to in the experiment. If a symbol has no story, the model is not finished.

ma t hbb P (X = k) = bin o mnk p^{k} (1 - p)^{n - k}

Now read the statement backwards: what would have to be known to use it? Backwards reading is often the difference between recognizing a formula and knowing when it applies.

ma t hbb P (X = k) = e^{- l amb d a} l amb d a^{k} / k! q u a d t e x t (P o i sso n)

Test the expression at an edge case or simple symmetric case. Probability formulas should survive sanity checks before you trust the arithmetic built on them.

Worked example ladder

A small experiment you can actually do

A five-step habit for every example in this chapter.

What usually goes wrong

When you notice this mistake, do not merely correct the final number. Return to the first line where the model became ambiguous. Probability errors are often representation errors wearing arithmetic clothing.

Questions beginners are right to ask

Binomial or geometric?

Binomial fixes the number of trials and counts successes; geometric fixes the number of successes at one and waits for the trial on which it occurs.

Why does Poisson have one parameter?

Its mean and variance are both lambda under the basic model, so one rate parameter controls location and spread.

What does memoryless mean?

Conditional on having waited already, the remaining waiting-time distribution is unchanged. This is a strong structural property, not ordinary forgetfulness.

Where the abstraction earns its keep

For each application, ask what counts as an outcome, what the model treats as random, and which assumptions are approximations. This is how probability becomes a modelling language instead of a catalogue of formulas.

Connections: do not store chapters in separate boxes

Problem-solving clinic: from recognition to fluency

There is a stage where every worked example looks clear but a fresh problem still feels foreign. The cure is not another formula; it is practice choosing the representation. Before equations, do a sixty-second scan: identify the experiment, what is known, what remains uncertain, the quantity being asked for, the assumption doing the heavy lifting, and one impossible answer that gives you a sanity bound.

The expert loop returns every calculation to the original story.

Case clinic A: Bernoulli and binomial

A Bernoulli variable records one yes/no trial. Summing n independent Bernoulli(p) variables produces Binomial(n,p), and the binomial coefficient counts which trials succeeded.

Case clinic B: Geometric waiting

The geometric distribution models the trial number of the first success under independent repeated trials with constant success probability. Its memoryless property is exceptional among discrete distributions.

Solve or reason about it twice: once exactly and once with a rough estimate, simulation, or symmetry argument. If the two approaches disagree dramatically, investigate before trusting the more sophisticated calculation.

Debug a confident wrong answer

Two questions to answer without notes

Binomial or geometric? Binomial fixes the number of trials and counts successes; geometric fixes the number of successes at one and waits for the trial on which it occurs.

Why does Poisson have one parameter? Its mean and variance are both lambda under the basic model, so one rate parameter controls location and spread.

A notebook protocol for proficiency

Give this chapter one notebook page divided into four quadrants: picture, formula, example, mistake. Redraw the main visual from memory, narrate one formula in English, invent a fresh story using the same mathematics, and record the most tempting wrong move. Revisit the page after two days and again after a week.

A mastery check before you move on

Try these without looking back. If one item feels slippery, return to the corresponding example and rebuild it rather than memorizing the answer.

Give a one-minute explanation of the chapter title to a curious teenager without a formula.
Invent a tiny example with at most six elementary outcomes and solve it completely by enumeration.
State one assumption that would make your example invalid and identify exactly which line would break.
Draw the mental map from memory and connect at least two boxes to an earlier or later chapter.
Write one question whose answer you still do not know. Good questions show that the concept has become active rather than passive.