Quick Summary: Probability & Statistics is the second-largest subject in GATE DA (~12–16 marks) and the mathematical foundation of Machine Learning, the largest. The official IIT Madras GATE 2027 syllabus runs from counting and probability axioms through Bayes’ theorem, expectation and correlation, nine named distributions, the Central Limit Theorem, confidence intervals, and the z, t and chi-squared tests. This guide covers every listed topic with diagrams, a worked Bayes example, and the exact patterns GATE uses to test them.
By Piyush Wairale — Instructor, BS Data Science program at IIT Madras · IIT Madras alumnus · 10,000+ GATE students mentored · Last updated: August 2026 · Verified against the official IIT Madras GATE 2027 DA syllabus
Key Takeaways
- Probability & Statistics contributes roughly 12–16 of GATE DA’s 85 core marks directly — and indirectly powers Machine Learning questions (naive Bayes, logistic regression, LDA) worth many more.
- Bayes’ theorem — P(A|B) = P(B|A)·P(A)/P(B) — is the single most-tested formula in the subject; expand the denominator with the law of total probability in every problem.
- For the binomial distribution, mean = np and variance = np(1−p); for Poisson, mean = variance = λ — the equality of Poisson’s mean and variance is a repeated one-mark fact.
- The Central Limit Theorem says the sample mean of n i.i.d. observations is approximately Normal(μ, σ²/n) for large n regardless of the population’s distribution — the bridge from probability to statistics.
- Choose the test by what you know: z-test when population variance is known (or n large), t-test when variance is estimated from a small sample, chi-squared for variance claims and goodness-of-fit/independence of categorical data.
On this page
Why P&S matters · Counting, axioms & conditional probability · Bayes’ theorem (worked example) · Random variables, expectation & correlation · The nine distributions · CLT & confidence intervals · z, t and chi-squared tests · How GATE tests P&S · 5-week plan · FAQs
If Machine Learning is the engine of GATE DA, Probability & Statistics is the fuel line. It is the second-largest subject by direct weightage, and it silently powers a third of the ML questions: naive Bayes is conditional probability, logistic regression is a probability model, LDA is Gaussian densities, and cross-validation is sampling theory. Prepare this subject well and two sections of the paper move at once. This article covers every topic named in the official IIT Madras GATE 2027 syllabus in teaching order, with the diagrams, formula tables and worked examples you need — and it pairs with the structured Probability & Statistics for GATE DA Course & Test Series if you want lectures, PYQs and tests on top.
Watch Free: Probability & Statistics Lectures on YouTube
Piyush Wairale’s channel has free subject-wise GATE DA lectures — including the Probability & Statistics series with GATE-style numericals solved on the whiteboard:
- All subject playlists — Probability & Statistics, Linear Algebra, DBMS, Python/DSA and more
- Machine Learning for GATE DA — full playlist (uses this subject constantly)
Subscribe to Piyush Wairale IITM on YouTube for new GATE DA lectures, PYQ solutions and strategy sessions.
Why Probability & Statistics Punches Above Its Weightage
On paper, this subject is worth 12–16 marks. In practice, it is worth more, for two reasons. First, cross-section leverage: a naive Bayes question in the ML section is a conditional-probability question; the reasoning-under-uncertainty block of AI is Bayes’ theorem on graphs; even regression numericals borrow covariance and correlation. Second, predictability: probability questions in GATE recycle a small set of shapes — a Bayes “diagnostic test” scenario, a binomial counting setup, an expectation computation, a confidence-interval or test-selection question. Once you have solved thirty PYQs, the thirty-first feels familiar. The syllabus divides cleanly into two halves — probability (counting through distributions) and statistics (CLT through hypothesis tests) — and this guide follows that order.
Foundations: Counting, Axioms and Conditional Probability
Counting: permutations and combinations
Almost every “probability of an event” question begins with counting. Permutations count ordered arrangements: P(n, r) = n!/(n−r)!. Combinations count unordered selections: C(n, r) = n!/(r!(n−r)!). The classical probability of an event is favourable outcomes over total outcomes, so most early questions reduce to two careful counts. Speed habits worth building: C(n, r) = C(n, n−r); the number of binary strings of length n is 2ⁿ; and the number of ways to arrange n items with repetitions n₁, n₂, … is n!/(n₁!n₂!…).
Axioms, events and independence
The three Kolmogorov axioms — P(A) ≥ 0, P(S) = 1, and additivity of mutually exclusive events — generate everything else you use: P(Aᶜ) = 1−P(A), and the inclusion-exclusion identity P(A∪B) = P(A) + P(B) − P(A∩B). Two definitions that GATE deliberately contrasts: events are mutually exclusive when P(A∩B) = 0 (they cannot co-occur) and independent when P(A∩B) = P(A)·P(B) (one tells you nothing about the other). Mutually exclusive events with non-zero probabilities are never independent — knowing one occurred guarantees the other didn’t. That exact statement has appeared as an MCQ option more than once.
Marginal, conditional and joint probability
Conditional probability is P(A|B) = P(A∩B)/P(B). From it come the two workhorses: the multiplication rule P(A∩B) = P(A|B)·P(B), and the law of total probability — if B₁,…,Bk partition the sample space, then P(A) = Σᵢ P(A|Bᵢ)·P(Bᵢ). Joint distributions over two variables are usually presented in GATE as a small table; marginals are row/column sums, and conditionals are cell over marginal. Practise reading these tables fast — they also appear in the DBMS-flavoured data questions and in ML.
Bayes’ Theorem: The Most-Tested Formula in the Paper
Bayes’ theorem inverts a conditional probability:
P(A | B) = P(B | A) · P(A) / P(B), with P(B) = P(B|A)·P(A) + P(B|Aᶜ)·P(Aᶜ)
Worked example (the classic GATE shape). A disease affects 1% of a population. A test detects it with probability 0.95 when present (sensitivity) and gives a false positive with probability 0.05 when absent. A random person tests positive — what is the probability they actually have the disease?
P(D|+) = P(+|D)·P(D) / [P(+|D)·P(D) + P(+|Dᶜ)·P(Dᶜ)] = (0.95 × 0.01) / (0.95 × 0.01 + 0.05 × 0.99) = 0.0095 / (0.0095 + 0.0495) = 0.0095/0.059 ≈ 0.161. Despite a “95% accurate” test, a positive result means only a 16% chance of disease — because the base rate is tiny. This base-rate effect is exactly what GATE wants you to internalise; the same computation reappears with spam filters, defective machines and sensor alarms. Notice it is also the same mathematics as the alarm-network posterior in our reasoning under uncertainty guide — one skill, two sections of the paper.
Random Variables, Expectation, Variance and Correlation
A random variable attaches a number to each outcome; discrete RVs have a probability mass function (PMF), continuous RVs a probability density function (PDF), and both have a cumulative distribution function F(x) = P(X ≤ x) — non-decreasing, from 0 to 1, with P(a < X ≤ b) = F(b) − F(a). For continuous variables, the density integrates to 1 and P(X = x) = 0 for any single point — a favourite true/false statement. The syllabus also names the conditional PDF, f(x|y) = f(x,y)/f(y): the same conditioning idea, done with densities.
The moment toolkit: E[X] = Σx·p(x) (or ∫x·f(x)dx), and Var(X) = E[X²] − (E[X])² — memorise this second form; it is the fastest route in almost every numerical. Linearity of expectation, E[aX + bY] = aE[X] + bE[Y], holds with or without independence — while Var(aX + b) = a²Var(X), and Var(X + Y) = Var(X) + Var(Y) only under independence (in general add 2Cov(X, Y)). Covariance Cov(X,Y) = E[XY] − E[X]E[Y] measures joint variation; correlation ρ = Cov(X,Y)/(σxσy) scales it to [−1, 1]. Two facts GATE probes: independence implies zero covariance, but zero covariance does not imply independence; and correlation measures only linear association. Finally, conditional expectation E[X|Y] and the tower rule E[E[X|Y]] = E[X] round out the syllabus list — the tower rule occasionally shows up as an elegant one-mark shortcut.
Mean, median and mode of a dataset — plus standard deviation — are listed explicitly and appear as quick data-interpretation questions: remember the median is robust to outliers while the mean is not, and for right-skewed data typically mode < median < mean.
The Nine Named Distributions
| Distribution | PMF / PDF | Mean | Variance |
|---|---|---|---|
| Discrete uniform {1..n} | 1/n | (n+1)/2 | (n²−1)/12 |
| Bernoulli(p) | pˣ(1−p)¹⁻ˣ | p | p(1−p) |
| Binomial(n, p) | C(n,k)pᵏ(1−p)ⁿ⁻ᵏ | np | np(1−p) |
| Poisson(λ) | e⁻ᵏλᵏ/k!* | λ | λ |
| Continuous uniform [a,b] | 1/(b−a) | (a+b)/2 | (b−a)²/12 |
| Exponential(λ) | λe⁻ᵏˣ* | 1/λ | 1/λ² |
| Normal(μ, σ²) | bell curve | μ | σ² |
| t (ν d.o.f.) | heavy-tailed bell | 0 | ν/(ν−2), ν>2 |
| Chi-squared (k d.o.f.) | sum of k squared Z’s | k | 2k |
*Exponents shown with λ: Poisson PMF is e−λλk/k!; exponential PDF is λe−λx for x ≥ 0.
Two memory anchors that convert directly into marks: the exponential distribution is memoryless — P(X > s+t | X > s) = P(X > t), the only continuous distribution with this property — and the Poisson approximates the binomial when n is large and p small with λ = np. The t-distribution has heavier tails than the normal and approaches it as degrees of freedom grow; the chi-squared arises as the distribution of a sum of squared standard normals, which is exactly why it appears in variance tests.
The Central Limit Theorem and Confidence Intervals
The Central Limit Theorem (CLT) is the hinge between the two halves of the syllabus. It states: if X₁,…,Xn are i.i.d. with mean μ and variance σ², then for large n the sample mean X̄ is approximately Normal(μ, σ²/n) — regardless of the shape of the original distribution. Three testable consequences: the standard deviation of the sample mean (the standard error) is σ/√n, so quadrupling the sample halves the error; sums and averages of enough i.i.d. anything look Gaussian; and this is precisely what justifies the z-test and confidence intervals below.
A confidence interval turns a point estimate into a range: X̄ ± zα/2·σ/√n for a (1−α) confidence level, with z = 1.96 for 95% (memorise also 1.645 for 90% and 2.576 for 99%). Interpretation matters for MCQs: a 95% CI means that the procedure captures the true mean in 95% of repeated samples — not that “the true mean has a 95% probability of being in this particular interval.” When σ is unknown and n is small, replace z with the t-critical value at n−1 degrees of freedom.
Hypothesis Testing: z-test, t-test and Chi-Squared Test
A hypothesis test asks whether the data are consistent with a null hypothesis H₀. You compute a test statistic, compare it with a critical value (or compute a p-value), and reject H₀ if the statistic lands in the tail. The GATE skill is mostly choosing the right test and computing the statistic:
| Test | Use when | Statistic | Reference distribution |
|---|---|---|---|
| z-test | Testing a mean, σ known (or n large) | z = (X̄−μ₀)/(σ/√n) | Standard normal |
| t-test | Testing a mean, σ unknown, small n | t = (X̄−μ₀)/(s/√n) | t with n−1 d.o.f. |
| Chi-squared test | Variance claims; goodness-of-fit; independence of categorical variables | Σ(O−E)²/E or (n−1)s²/σ₀² | Chi-squared |
Vocabulary GATE expects: a Type I error rejects a true H₀ (its probability is the significance level α); a Type II error fails to reject a false H₀; the p-value is the probability, under H₀, of a statistic at least as extreme as observed — reject when p < α. One- vs two-tailed matters: “mean has changed” is two-tailed; “mean has increased” is one-tailed, and the critical z at 5% shifts from 1.96 to 1.645 accordingly.
How GATE Actually Tests Probability & Statistics
- Bayes diagnostic scenarios (2 marks, NAT): the disease-test computation above, reskinned as spam filters, defective factories or sensor alarms. Always expand the denominator via total probability.
- Counting-based probability (1–2 marks): cards, dice, balls-in-urns; the work is in the combinatorics, not the probability.
- Expectation/variance numericals (2 marks): compute E[X] and Var(X) from a small PMF table, or use Var = E[X²] − (E[X])² after a moment computation.
- Distribution recognition (1 mark): “arrivals per minute” → Poisson; “time until failure” → exponential; “successes in n trials” → binomial; memoryless → exponential; mean = variance → Poisson.
- CLT / standard error (1–2 marks): how does the standard error change with n; probability that a sample mean exceeds a threshold after standardising.
- Test selection and interpretation (1–2 marks): pick z vs t vs chi-squared from the scenario; interpret p-values and Type I/II errors; compute a z or t statistic.
- Correlation conceptuals (1 mark): zero correlation vs independence; effect of linear transformations on ρ (unchanged in magnitude).
The 5-Week Probability & Statistics Study Plan
- Week 1 — Foundations: counting drills, axioms, conditional probability, independence vs mutual exclusivity; 25 mixed problems.
- Week 2 — Bayes and random variables: ten Bayes scenarios end-to-end; expectation/variance numericals; covariance and correlation.
- Week 3 — Distributions: build the mean/variance table above from memory; distribution-recognition drills; Poisson↔binomial and exponential memorylessness.
- Week 4 — Statistics: CLT standardisation problems, confidence intervals at all three standard levels, z/t/chi-squared selection and computation.
- Week 5 — Integration: full sectional tests, GATE DA + GATE CS probability PYQs, error-log review; then move straight into Machine Learning, which uses everything you just built.
Study Probability & Statistics the Structured Way
Every topic above — with derivations, solved GATE PYQs, topic-wise tests and doubt support:
- Probability & Statistics for GATE DA Course & Test Series — the complete syllabus, taught by IIT Madras alumnus Piyush Wairale.
- Complete GATE DA 2027 Course — all 7 subjects, 250+ hours, 4,000+ questions and full-length AI mock tests in one package.
FAQs on Probability & Statistics for GATE DA
How many marks does Probability & Statistics carry in GATE DA?
Typically 12–16 of the 85 core marks — second only to Machine Learning. Counting the ML questions built on probability (naive Bayes, logistic regression, LDA), its effective weight is considerably higher.
Should I study this before Machine Learning?
Yes — it is the single most important prerequisite. The recommended order in our 6-month plan is Linear Algebra and Probability & Statistics first, then Machine Learning.
Which distributions must I memorise with mean and variance?
All nine in the table above: discrete uniform, Bernoulli, binomial, Poisson, continuous uniform, exponential, normal/standard normal, t and chi-squared. The most-tested single facts: binomial np and np(1−p); Poisson mean = variance = λ; exponential memorylessness.
When do I use a t-test instead of a z-test?
Use t when the population standard deviation is unknown and estimated from a small sample (statistic uses s, reference distribution is t with n−1 degrees of freedom). With σ known or n large, use z.
Which book should I follow?
Sheldon Ross — “A First Course in Probability” for the probability half and “Introduction to Probability and Statistics for Engineers and Scientists” for the testing half. Full recommendations in the best books for GATE DA guide.
Probability & Statistics is the best drilling investment in GATE DA: a compact formula set, question shapes that repeat, and direct leverage into Machine Learning and AI. Build the distribution table into muscle memory, make Bayes computations automatic, and learn to pick the right test in ten seconds — the marks follow. For the full subject roadmap, see the GATE DA Syllabus 2027 breakdown.
Master GATE DA Probability & Statistics
From counting to chi-squared — lectures, solved PYQs, sectional tests and mentorship by Piyush Wairale (IIT Madras).
Join the P&S Course & Test Series Get the Complete GATE DA CourseRecent Post
Linear Algebra carries ~8–10 marks in GATE DA and powers PCA, regression and SVD in Machine Learning. This complete guide covers the full official syllabus — vector spaces, special matrices, quadratic forms, systems of equations, Gaussian elimination, eigenvalues, rank-nullity, LU decomposition and SVD — with diagrams, a solved eigenvalue example, GATE patterns and FAQs, by IIT Madras instructor Piyush Wairale.
Machine Learning is the highest-weightage subject in GATE DA (~18–22 marks). This complete guide covers the full official syllabus — regression, classifiers, SVM, decision trees, bias-variance, cross-validation, neural networks, clustering and PCA — with diagrams, GATE question patterns, a 6-week plan and FAQs, by IIT Madras instructor Piyush Wairale.
Master reasoning under uncertainty for GATE DA 2027 — Bayesian networks, conditional independence, d-separation, variable elimination with a fully worked alarm-network example, and sampling-based approximate inference. With diagrams, GATE question patterns, a 10-day plan and FAQs, by IIT Madras instructor Piyush Wairale.
Complete GATE RA 2027 guide to servo motors: closed-loop principle, AC/DC/BLDC types, continuous vs intermittent torque-speed zones, nested PID control, sizing, diagrams and worked example.
Complete GATE RA 2027 guide to stepper motors: VR/PM/hybrid types, step angle, full/half/microstepping, pull-in and pull-out torque-speed characteristics, drives, diagrams and worked examples.
Complete GATE RA 2027 guide to DC motors: principle of operation, back-EMF, torque equation, shunt/series/compound torque-speed characteristics, speed control, performance, diagrams and worked examples.
Learn Daily, Wherever You Are
Free lectures, exam updates, PYQ discussions, and job alerts — delivered through our YouTube channel and Telegram communities.
