Statistics › Section C · Probability › Topic C.1
Probability terminology
Probability turns "how likely?" into a number between 0 and 1. This topic sets up the vocabulary every later probability topic stands on: experiments, sample spaces, events, and the long-run meaning of chance.
Putting a number on chance
Statistics leans on probability at every turn: samples are drawn by chance, so their behaviour follows the mathematics of chance. Probability answers "how likely?" with a number between 0 and 1, grounded in the long run - the proportion of times an event would occur if the random process were repeated over and over. This topic sets up the vocabulary the whole probability chapter uses: experiments, outcomes, sample spaces and events.
Experiments, outcomes, sample spaces
A probability experiment is any process with an uncertain result: rolling a die, drawing a marble, surveying a randomly chosen person. One result of the experiment is an outcome. The sample space \(S\) is the set of every possible outcome: for one die, \(S = \{1, 2, 3, 4, 5, 6\}\); for one coin, \(S = \{H, T\}\). The word "random" matters: the outcome must be unpredictable in advance for the mathematics to apply.
Events
An event is any subset of the sample space - any collection of outcomes we care about. Rolling a die, the event "even" is \(\{2, 4, 6\}\), and the event "at least five" is \(\{5, 6\}\). Events get capital letters: \(A\), \(B\), \(T\). An event can hold one outcome (roll a two), several (roll an even number), or the whole sample space (roll anything at all). Wording matters throughout this chapter: the first step in any probability problem is reading carefully enough to name what the event actually is.
Equally likely outcomes
When every outcome has the same chance, probability becomes counting:
A fair die has six equally likely outcomes, so \(P(\text{even}) = \frac{3}{6}\) and \(P(\text{at least five}) = \frac{2}{6}\). The formula demands both halves: the outcomes must be equally likely (a weighted die breaks it), and every outcome must appear in the sample space exactly once. Probabilities always land between 0 (impossible) and 1 (certain), and the probabilities of all outcomes together total 1.
Theoretical versus experimental probability
There are two ways to obtain a probability. The theoretical way reasons from the experiment's structure: one face of six, so \(1/6\). The experimental (empirical) way observes data: flip a coin 2,000 times and divide the number of heads by 2,000. As trials accumulate, a careful experiment tends to estimate the theoretical value more closely, which is why the two methods agree in the long run. When a process is too complex to analyse - a game of solitaire, a busy queue - the experimental route, or a simulation that imitates the process many times, may be the only one available.
The law of large numbers
As repetitions grow, the relative frequency of an event settles toward its probability - the law of large numbers. Flip a fair coin many times and the proportion of heads drifts toward 0.5. What the law does not say is that short runs balance: after five straight heads, the next flip is still exactly 0.5, because the coin has no memory. Believing that past outcomes change the chance of independent future ones is the gambler's fallacy. Probability describes the long run; streaks are built into it.
AND, OR, complements, and "given"
Probability has its own connective words, and their meanings are precise. \(A\) AND \(B\): both events occur - the overlap of the two sets. \(A\) OR \(B\): at least one occurs, including both (unlike the everyday either-or). The complement, written \(A'\) and read not-\(A\): the event that \(A\) fails to happen. And \(A\) GIVEN \(B\), written \(P(A|B)\): the probability of \(A\) once \(B\) is known to have happened - a conditional question, developed fully in later topics. Reading a problem means finding these words and naming the events behind them.
Worked example: a jar of marbles
A jar holds 4 green, 3 yellow and 5 purple marbles, all the same size. One is drawn at random. 1. By colour the sample space is \(\{\text{green, yellow, purple}\}\), but the equally likely outcomes are the 12 individual marbles. 2. \(P(\text{green}) = \frac{4}{12} = \frac{1}{3}\). 3. \(P(\text{yellow}) = \frac{3}{12} = \frac{1}{4}\). 4. \(P(\text{purple}) = \frac{5}{12}\). 5. Check: the three probabilities total \(\frac{4}{12} + \frac{3}{12} + \frac{5}{12} = \frac{12}{12} = 1\), as they must - drawing some colour is certain.
Worked example: two dice, 36 outcomes
Roll two fair dice and add the faces. 1. Sample space: each die has 6 outcomes, so the experiment has \(6 \times 6 = 36\) equally likely pairs - written as (first die, second die), so (4, 3) and (3, 4) both count. 2. Sum of 7: the pairs (1,6), (2,5), (3,4), (4,3), (5,2) and (6,1) - six of them - give \(P(7) = \frac{6}{36} = \frac{1}{6}\). 3. Sum of 11: only (5,6) and (6,5), so \(P(11) = \frac{2}{36} = \frac{1}{18}\). 4. Check: no other sum has as many as six pairs, so 7 is the most likely total; sums near 2 or 12 have very few.
Common mistakes to avoid
First, treating outcomes as equally likely without justification - a weighted die or a trick coin breaks the counting formula. Second, forgetting order in two-stage experiments: (4,3) and (3,4) are different outcomes. Third, reading "or" as exclusive: in probability, \(A\) OR \(B\) includes both happening. Fourth, applying the law of large numbers to short runs - streaks are not "due" to end. Fifth, letting the gambler's fallacy steer: five heads in a row make tails feel owed, but the next flip is untouched. Sixth, skipping the reading step: misnaming the event wastes every calculation after it.
Key takeaways
A probability experiment has outcomes; the sample space \(S\) collects them all; an event is a subset of \(S\). Probabilities lie between 0 and 1 and, across all outcomes, total 1. With equally likely outcomes, \(P(A)\) counts \(A\)'s outcomes over all of \(S\)'s. Theoretical probability reasons from structure, experimental probability from data, and the law of large numbers brings them together in the long run - never through short-run balancing. AND means overlap, OR means at least one, a complement means not-A, and "given" signals a conditional question.
Flashcards
Flashcard
A process with an unpredictable result
Test your understanding
Question 1 of 5
Flipping two fair coins, the sample space is:
Question 2 of 5
A bag holds 2 red, 6 blue and 4 white counters of the same size. One is drawn at random. What is P(blue)?
Question 3 of 5
After five heads in a row, the probability that the next fair-coin flip is heads is:
Question 4 of 5
In probability, "A OR B" means:
Question 5 of 5
Rolling two fair dice, what is the probability the sum is 8?
