Statistics and probability
19 lessons · 8 HL only · 147 past-paper questions in the AI HL bank
● SL AND HL
11 lessonsStatistics is only as good as its data. You learn how a sample stands in for a population, how the way it is chosen can bias it, and how to handle missing or wrongly recorded values. You also find an outlier with the 1.5 × IQR rule and decide whether to keep it.
A long list of numbers says little until you organise it. You sort data into frequency tables, draw histograms, read the median, quartiles and percentiles off a cumulative frequency graph, and use box and whisker diagrams to compare two data sets and judge whether data may be normal.
The mean, median and mode describe the centre of a data set; the IQR and standard deviation describe its spread. You find them with your GDC, estimate the mean of grouped data, and see what shifting or scaling the data does.
Pairs of measurements, such as a café's distance from a station and its daily customers, are bivariate data. You describe how the two move together, measure a straight-line link with Pearson's r, and predict with the regression line of y on x. You also learn when a prediction can't be trusted.
Probability measures how likely an event is, on a scale from 0 to 1. When the outcomes are equally likely you count them: P(A) = n(A)/n(U). When you can't count, you estimate from data, and either way you can predict how often the event should happen in many trials.
Venn diagrams, tree diagrams and tables turn a wordy probability question into regions and branches you can read off, often with no formula at all. A few rules cover the rest: the addition rule for "or", mutually exclusive events, conditional probability for "given that", and multiplying for independent events.
A random variable is a number decided by chance, such as the number of cars a household owns. Its probability distribution gives each value with its probability, and the probabilities add up to 1. The expected value E(X) is the long-run average: it predicts totals, sets fair prices and tells you whether a game is fair.
A binomial distribution counts how many times something happens in a fixed number of independent tries, each with the same chance. You write X ~ B(n, p), find its probabilities on your GDC, and use np for the number you expect.
Heights, masses and many other measurements are often modelled by the normal distribution: a symmetric bell-shaped curve centred on the mean. Its probabilities are areas under the curve. You estimate them with the 68–95–99.7 rule, find them with your GDC, and run the GDC backwards to find a value from a probability.
Spearman's rank correlation coefficient, rₛ, measures how well two rankings agree. You rank each variable, then your GDC finds rₛ from the ranks. Use it when the data are ranks, when the link is curved but always rising or always falling, or when an outlier would distort Pearson's r.
A hypothesis test asks whether data give real evidence for a claim or could be chance. You set up hypotheses, use p-values and significance levels, and run three GDC tests: χ² for independence, χ² goodness of fit and the t-test for two means.
◆ HL ONLY
8 lessonsGood statistics start before any calculation: honest, precise questions, the right variables and the right data. You also group numerical data so a χ² test works, adjust the degrees of freedom when you estimate parameters, and judge whether a method is reliable and valid.
Real data often follow a curve, not a line. Your GDC finds the best curve of the type you choose, and you judge the fit with the sum of square residuals and R². A high R² never settles the choice alone: the context matters too.
Scaling, shifting and adding random variables changes their mean and spread in predictable ways. For a linear combination, expected values always add. For independent X and Y, Var(aX ± bY) = a²Var(X) + b²Var(Y), even for a difference. From a random sample, x̄ and s_(n−1)² are unbiased estimates of the population mean and variance.
Add or subtract independent normal variables and the result is normal too. In particular, the mean X̄ of a random sample of size n from N(μ, σ²) is N(μ, σ²/n). The central limit theorem says that for a large sample X̄ is approximately normal whatever the population looks like, and in exams n > 30 is large enough.
A confidence interval is a range of plausible values for a population mean, worked out from one sample. If σ is known you use the normal distribution; if not, the t-distribution. Then you say what the interval means in context.
A Poisson distribution counts events in a fixed stretch of time or space, when they happen independently at a steady average rate. Its mean and variance are both m, your GDC gives its probabilities, and independent Poisson counts add to another Poisson count.
HL adds tests for a mean (z or t), a proportion (binomial), a Poisson rate and correlation (ρ = 0). You also find critical regions and the probabilities of Type I and Type II errors.
A Markov chain moves between states, such as shops a customer uses, where the next step depends only on the current state. A transition matrix T holds the step probabilities, and sₙ = Tⁿs₀ predicts the state after n steps. The steady state, an eigenvector of T for the eigenvalue 1, gives the long-term proportions.