Chapter 1 · Introduction

Chapter 1 — Probability

From data to models: how to summarise what we observe, how to build a probability space that explains it, and how random variables, expectations and joint distributions turn chance into mathematics.

Statistics begins with data: the waiting times at a clinic, the yields of a crop variety on forty plots, the daily returns of a stock. Before modelling anything we must look at the data. That means sorting them into a frequency table, drawing a histogram and a box plot, and condensing them into a few numbers that describe their centre, spread, asymmetry and the strength of linear relations between variables. These descriptive summaries are the vocabulary of the whole subject. Every later estimator, test and model is, in the end, a statement about some such summary computed from a random sample.

To say anything about data that we have not yet seen, we need a model of chance. Kolmogorov's 1933 axioms supply one: a sample space, a σ-field of events and a countably additive probability measure. From three axioms come inclusion–exclusion, conditional probability, the law of total probability and Bayes' theorem, which is the rule for updating beliefs on evidence and the reason a positive test for a rare disease is less alarming than it looks. Random variables then turn outcomes into numbers. Their distribution functions, expectations, moments, and joint and conditional distributions are the language in which the rest of the course (limit theorems, Markov chains, estimation, testing, linear models) is written.

Data & graphs→Location & spread→Moments & correlation→σ-fields & axioms→Conditioning & Bayes→Independence→Random variables→Expectation & mgf→Joint & conditional laws
🎯 By the end of this chapter you can…
  • build frequency tables, histograms (with unequal classes), stem-and-leaf displays and box plots, and identify outliers;
  • compute and interpret the mean, median, mode, trimmed mean, variance, IQR, mean deviation, CV, moments, skewness and kurtosis, for raw and for grouped data, and prove their extremal properties and inequalities;
  • compute a correlation coefficient and the two least-squares lines, and use $-1\le r\le1$ and $b_{yx}b_{xy}=r^2$;
  • work with σ-fields and the axioms of probability, inclusion–exclusion, classical and geometric probability, conditional probability, total probability and Bayes' theorem;
  • distinguish pairwise from mutual independence and apply the Borel–Cantelli lemmas, the birthday, matching, Pólya urn and gambler's ruin problems;
  • handle distribution functions, densities, expectations, moment generating functions, joint, marginal and conditional distributions, conditional expectation, Jacobian transformations and the bivariate normal.
Teaser: From Frequencies to Probabilities with Two Dice

Roll two fair dice again and again and record the total. The blue bars are the relative frequencies of the totals $2,\dots,12$ (descriptive statistics, §1.1). The red dots are the classical probabilities $\tfrac{6-|s-7|}{36}$ obtained by counting the 36 equally likely outcomes (§1.2). Press Play and watch the data approach the model. The readouts compare the sample mean and variance with $E[S]=7$ and $\operatorname{Var}(S)=\tfrac{35}{6}$ (§1.3).

rolls—
sample mean (7)—
sample variance (5.8333)—
largest gap |freq − prob|—

How this is examined

This chapter is the foundation of the Unit 4 questions, and it is examined directly and indirectly. Part B questions often ask for one number. Typical requests are the mean or variance of transformed or combined data, the correlation coefficient determined by two given regression lines, the value of $a$ minimising $\sum|x_i-a|$, a probability from inclusion–exclusion or a geometric picture, a posterior probability from Bayes' theorem, the number of sets in a generated σ-field, or the expectation of a simple random variable. Part C, where more than one option may be correct, prefers statements to be judged. Is $r$ invariant under change of origin and scale? Must $\beta_2\ge1$? Does pairwise independence imply mutual independence? Do disjoint events with positive probabilities ever become independent? Which of these functions are distribution functions? Does zero correlation imply independence? Every wrong option is built on one of the counterexamples collected in the ⚠️ boxes of the four sections. GATE ST asks similar questions in numerical-answer and multiple-select formats.

Sections in this chapter

✅ Before you start

You need school algebra and permutations and combinations for §§1.1–1.2. You need BSc calculus for §§1.3–1.4: improper integrals, double integrals with a change of variables (the Jacobian), and convergent series such as $\sum x^k/k!$ and $\sum kx^{k-1}$. Set language (unions, intersections, complements, De Morgan's laws, countable and uncountable sets) is used from §1.2 on. If it is rusty, revise it from the first chapter of the sister course CSIR NET Unit 1: Analysis and Linear Algebra, which also covers the Cauchy–Schwarz inequality used in §1.1. If this is your first course of this kind, read the welcome page to see how the sections are organised and how to use the 🎯 Exam corner.