Statistics › Section D · Discrete random variables › Topic D.1

Probability distributions for discrete variables

A random variable turns the outcomes of chance into numbers, so that averages and spreads of random results become computable. This topic builds the probability distribution - the table that says what the variable can be, and how likely each value is.

Outcomes become numbers

Dice show faces, coins show heads and tails, but the questions we care about are numeric: how many heads, how many customers, how long a wait. A random variable assigns a number to every outcome of a chance process, so averages, spreads and whole distributions become computable. This chapter builds the tools for discrete random variables - the distribution that lists what can happen, then the expected value that summarises it - starting with the distribution.

Random variables: discrete and continuous

A random variable \(X\) attaches one number to each outcome: the number of heads in two flips, the sum of two dice, the number of green marbles drawn. Discrete random variables take listable, separate values - counts, sums, ratings - usually whole numbers. Continuous random variables take any value in an interval - times, weights, temperatures - and their probabilities behave differently: they are areas under a curve, and any single exact value has probability 0. This topic works with discrete variables; the continuous case gets its own treatment in a later topic.

The probability distribution

A discrete random variable is described by its probability distribution: a list of every value it can take, each with its probability. Written \(P(X = x)\), it is a table with two columns - the value \(x\) and the probability \(P(x)\). Many texts call it a probability mass function, keeping the word density for continuous variables. Two properties make a list a distribution: every probability lies between 0 and 1, and they sum to exactly 1. The sum rule doubles as a self-check: it says the experiment must land somewhere, so all the listed chances together are certainty.

Building a distribution from an experiment

Distributions are built from sample spaces by grouping outcomes that share a value. Let \(X\) be the number of green marbles in two draws from the jar of 12 (4 green, 8 other), without replacement - the two-draw tree from the tree-diagram topic. Grouping the four paths by green count: \(P(X = 0) = \frac{56}{132} = \frac{14}{33}\) (both other), \(P(X = 1) = \frac{64}{132} = \frac{16}{33}\) (one of each, two paths), \(P(X = 2) = \frac{12}{132} = \frac{1}{11}\) (both green). The three probabilities total \(\frac{56+64+12}{132} = 1\), as they must.

Reading probabilities from a distribution

A distribution answers more than point questions. \(P(X = 1) = \frac{16}{33}\) reads straight off the table. \(P(X \ge 1)\) adds the qualifying entries: \(\frac{64+12}{132} = \frac{76}{132} = \frac{19}{33}\). \(P(X \le 1) = \frac{56+64}{132} = \frac{120}{132} = \frac{10}{11}\). And complements work as always: \(P(X \ge 1) = 1 - P(X = 0) = 1 - \frac{14}{33} = \frac{19}{33}\), agreeing with the direct addition. Every table question is either one entry, a sum of entries, or one minus something.

Worked example: is this a distribution?

Two proposed lists for the number of rainy days in a two-day weekend, so the possible values are 0, 1 and 2. Candidate A: \(P(0) = 0.3\), \(P(1) = 0.45\), \(P(2) = 0.25\). Candidate B: \(P(0) = 0.2\), \(P(1) = 0.5\), \(P(2) = 0.4\). 1. Range check: every entry in both candidates lies between 0 and 1, so both pass so far. 2. Sum check: A totals \(0.3 + 0.45 + 0.25 = 1.00\) - valid. B totals \(0.2 + 0.5 + 0.4 = 1.1\) - invalid, since no experiment can exceed certainty. 3. Check: a valid list can still be a poor model, but an invalid one cannot be a model at all. The two properties are necessary, not sufficient for truth.

Worked example: sixes in two rolls

Let \(X\) be the number of sixes when a fair die is rolled twice. 1. Sort the 36 equally likely pairs by their six count: no six in \(5 \times 5 = 25\) pairs, exactly one six in \(5 + 5 = 10\) pairs (six first or six second), and two sixes in 1 pair. 2. The distribution: \(P(X = 0) = \frac{25}{36}\), \(P(X = 1) = \frac{10}{36}\), \(P(X = 2) = \frac{1}{36}\). 3. \(P(X \ge 1) = 1 - \frac{25}{36} = \frac{11}{36}\). 4. \(P(X \le 1) = \frac{25 + 10}{36} = \frac{35}{36}\). Check: \(\frac{25 + 10 + 1}{36} = 1\), and the complement rule agrees, since \(1 - \left(\frac{5}{6}\right)^2 = \frac{11}{36}\).

The distribution as a picture

A distribution draws as a histogram with one bar per value: value on the horizontal axis, probability on the vertical. The green-count bars sit at heights \(\frac{14}{33} \approx 0.42\), \(\frac{16}{33} \approx 0.48\) and \(\frac{1}{11} \approx 0.09\), and the heights total exactly 1. The picture shows shape at a glance: one green is the likeliest count and two greens are rare. It also fixes notation - capital \(X\) for the random variable itself, lowercase \(x\) for a particular value it might take.

Common mistakes to avoid

First, accepting a table whose probabilities do not sum to 1 - run the check before using it. Second, mixing values with probabilities: 2 is a count of greens, and \(\frac{1}{11}\) is its probability. Third, reading \(P(X \ge 1)\) as \(P(X = 1)\): "at least" spans table entries. Fourth, forgetting that different outcomes can share a value - the two one-of-each paths merge into a single \(P(X = 1)\). Fifth, expecting the tallest bar to be the average; that idea arrives with expected value in the next topic. Sixth, leaving out a value the variable can take, such as 3 rainy days in a three-day weekend.

Key takeaways

A random variable \(X\) assigns a number to each outcome; discrete variables take listable values, continuous ones take intervals. A probability distribution lists each value \(x\) with \(P(X = x)\), valid only if every probability lies between 0 and 1 and the total is exactly 1. Distributions are built by grouping sample-space outcomes that share a value - the jar's four two-draw paths collapse to three green counts. Table questions are point, range or complement sums: \(\frac{16}{33}\), \(\frac{19}{33}\), \(1 - \frac{14}{33}\). One bar per value draws the distribution, with bar heights totalling 1.

Flashcards

Flashcard

Random variable (X)

A number assigned to each outcome of a random process

Outcomes become arithmetic.

Test your understanding

Question 1 of 5

Which property must every probability distribution satisfy?

Question 2 of 5

A proposed distribution lists P(0) = 0.2, P(1) = 0.5 and P(2) = 0.4. It is:

Question 3 of 5

X has P(0) = 0.15, P(1) = 0.35, P(2) = 0.30 and P(3) = 0.20. What is P(X ≥ 2)?

Question 4 of 5

A distribution lists P(0) = 0.1, P(1) = 0.3, P(2) = 0.4 and P(3) = k. What must k be?

Question 5 of 5

A random variable is best described as:

← Previous topicNext topic →