LAISHA
Open Laisha

IIT JAM Statistical Inference I: Estimation: practice questions

IIT JAM - Mathematical Statistics (MS) · India · IITs (JAM 2026 organised by IIT Bombay)

Original practice questions written to the published syllabus · not past papers
10worked here, free
84written to this exam in this subject
11syllabus units

About IIT JAM

IIT JAM - Mathematical Statistics (MS) is a postgraduate admission exam run by IITs (JAM 2026 organised by IIT Bombay), for candidates in India. Computer-based, 3 hours, 60 questions, 100 marks, English. Section A 30 MCQs (10 x 1 mark, 20 x 2 marks); Section B 10 MSQs (2 marks, no negative or partial marking); Section C 20 NAT (10 x 1, 10 x 2, no negative). MCQ wrong answers lose 1/3 or 2/3 mark.

Source: jam2026.iitb.ac.in.

Checked against the exam body's published information. Formats and fees change; confirm the current edition there before you book.

Statistical Inference I: Estimation in IIT JAM

Numerical and multiple-select items: UMVUEs of functions of parameters, Fisher information and CRLB values, MLEs including uniform and non-regular cases, invariance, consistency checks, and pivotal confidence intervals.

How to estimate well and know it: sufficiency and completeness, unbiasedness and minimum variance, the Cramer-Rao bound, the Rao-Blackwell and Lehmann-Scheffe theorems, the methods of moments, maximum likelihood and minimum chi-square, and confidence intervals.

Our 84 IIT JAM questions in this subject: 32 easy, 38 medium, 14 hard, across 11 units.

10 practice questions, worked

Choose an answer, then open the working. Each explanation says why the right answer is right and why the others are not.

These 10 were written to IIT JAM and are not shown on any other exam's page.

Question 1 · hard

Five quadrat counts are iid Poisson(λ) with total T = 12. What is the UMVUE estimate of λ²?

  1. 5.76
  2. 4.80
  3. 5.28
  4. 6.60
Answer and working

C. 5.28

T ~ Poisson(nλ), so E[T(T − 1)] = (nλ)², and T(T − 1)/n² = 132/25 = 5.28 is unbiased and a function of T. The plug-in X̄² = 5.76 is biased upward by λ/n; dividing by n(n − 1) applies the binomial formula wrongly.

Question 2 · medium

When does an unbiased estimator T of ψ(θ) attain the Cramér-Rao lower bound?

  1. Iff the support of the distribution depends on θ
  2. Iff the score is linear in T: ∂ log L/∂θ = k(θ)[T − ψ(θ)]
  3. Whenever T is the maximum-likelihood estimator of ψ(θ)
  4. Whenever T is a function of a complete sufficient statistic
Answer and working

B. Iff the score is linear in T: ∂ log L/∂θ = k(θ)[T − ψ(θ)]

Equality in the Cauchy-Schwarz step requires the score to be a linear function of T; this happens only in one-parameter exponential families for ψ(θ) = E(T). A UMVUE need not attain the bound (s² does not), and support depending on θ breaks the regularity conditions altogether.

Question 3 · easy

Why does the Cramér-Rao inequality not apply to estimating θ from uniform (0, θ) samples?

  1. The support depends on θ, violating regularity
  2. No unbiased estimator of θ exists for the uniform distribution
  3. The Fisher information is infinite at every value of θ
  4. The uniform has no finite variance, so no bound can be computed
Answer and working

A. The support depends on θ, violating regularity

Differentiation under the integral sign fails when the limits involve θ. Indeed (n + 1)X₍ₙ₎/n has variance θ²/[n(n + 2)], of order 1/n², beating any 1/n 'bound'. The uniform has finite variance, and unbiased estimators of θ exist.

Question 4 · hard

Counts are iid Poisson(λ). What is the Cramér-Rao lower bound for an unbiased estimator of ψ(λ) = e^(−λ), at λ = 2 with n = 10?

  1. 0.0009
  2. 0.2000
  3. 0.0271
  4. 0.0037
Answer and working

D. 0.0037

Bound = [ψ′(λ)]²/[nI(λ)] = e^(−2λ)/(n/λ) = λe^(−2λ)/n = 2e^(−4)/10 = 0.0037. 0.2000 is the bound for λ itself; omitting the square of ψ′ gives 0.0271; dividing by λ instead of multiplying gives 0.0009.

Question 5 · easy

In practice, what is the most common modern use of method-of-moments estimates in models fitted by maximum likelihood?

  1. As final estimates, because they are always more efficient
  2. As a formal test of whether the maximum-likelihood estimate is unbiased
  3. As starting values for maximum-likelihood iterations
  4. As standard errors, by taking the square root of the moments
Answer and working

C. As starting values for maximum-likelihood iterations

Moment estimates are cheap, consistent and usually close enough for Newton-Raphson or scoring to converge from them. They are not more efficient, and they do not give standard errors directly.

Question 6 · medium

Why does the modified minimum chi-square method fail when an observed class frequency is zero?

  1. Its criterion divides by the observed frequency, so an empty class means division by zero
  2. The degrees of freedom become negative when any class is empty
  3. Its criterion divides by the expected frequency, which then becomes zero as well
  4. A zero frequency makes the likelihood identically zero for every θ
Answer and working

A. Its criterion divides by the observed frequency, so an empty class means division by zero

Neyman's criterion is Σ(O − E)²/O. Empty classes must be pooled first or Pearson's version (division by E) or maximum likelihood used instead.

Question 7 · medium

Ten plots give s² = 0.25 (t/ha)². With χ²₉ points 19.023 (upper 2.5%) and 2.700 (lower 2.5%), what is the 95% confidence interval for σ²?

  1. 0.118 to 0.833
  2. 0.075 to 0.528
  3. 0.095 to 0.405
  4. 0.131 to 0.926
Answer and working

A. 0.118 to 0.833

(n − 1)s²/σ² ~ χ²₉, so σ² lies in [9 × 0.25/19.023, 9 × 0.25/2.700] = [0.118, 0.833]. Multiplying by the chi-square points inverts the pivot wrongly; using n = 10 instead of n − 1 = 9 shifts both limits.

Question 8 · medium

Lifetimes X₁, ..., Xₙ are iid exponential with rate λ. Which quantity is a pivot for building an exact confidence interval for λ?

  1. √n(X̄ − λ)/X̄, which has Student's t distribution with n − 1 df
  2. λΣXᵢ, which has a chi-square distribution with n degrees of freedom exactly
  3. ΣXᵢ/λ, which has a gamma distribution with shape n and rate 1
  4. 2λΣXᵢ, which has a chi-square distribution with 2n degrees of freedom
Answer and working

D. 2λΣXᵢ, which has a chi-square distribution with 2n degrees of freedom

λXᵢ ~ exponential(1), so λΣXᵢ ~ gamma(n, 1) and 2λΣXᵢ ~ χ²₂ₙ. ΣXᵢ/λ is not free of λ in distribution; X̄ estimates 1/λ, not λ, and t pivots need normal data.

Question 9 · easy

What is the Newton-Raphson update for maximising a log-likelihood ℓ(θ) of one parameter?

  1. θ₍ₖ₊₁₎ = θ₍ₖ₎ − ℓ(θ₍ₖ₎)/ℓ′(θ₍ₖ₎)
  2. θ₍ₖ₊₁₎ = θ₍ₖ₎ + ℓ″(θ₍ₖ₎)/ℓ′(θ₍ₖ₎)
  3. θ₍ₖ₊₁₎ = θ₍ₖ₎ − ℓ′(θ₍ₖ₎) / ℓ″(θ₍ₖ₎)
  4. θ₍ₖ₊₁₎ = θ₍ₖ₎ + ℓ′(θ₍ₖ₎)/ℓ″(θ₍ₖ₎)
Answer and working

C. θ₍ₖ₊₁₎ = θ₍ₖ₎ − ℓ′(θ₍ₖ₎) / ℓ″(θ₍ₖ₎)

Expanding ℓ′(θ) ≈ ℓ′(θₖ) + (θ − θₖ)ℓ″(θₖ) = 0 gives the update with a minus sign; near a maximum ℓ″ < 0, so the step moves in the direction of the score. θ − ℓ/ℓ′ solves ℓ = 0, the wrong equation.

Question 10 · medium

Ten Poisson counts total 30, so ℓ′(λ) = 30/λ − 10 and ℓ″(λ) = −30/λ². Starting from λ₀ = 2, what is one Newton-Raphson step?

  1. 2.6667
  2. 3.0000
  3. 1.3333
  4. 1.5000
Answer and working

A. 2.6667

ℓ′(2) = 15 − 10 = 5, ℓ″(2) = −7.5, so λ₁ = 2 − 5/(−7.5) = 2.6667. 3.0000 is the Fisher-scoring step (nI(2) = 10/2 = 5, λ₁ = 2 + 5/5), which happens to hit the MLE X̄ = 3 at once; adding instead of subtracting gives 1.3333.

A study plan for Statistical Inference I: Estimation in IIT JAM

This plan follows where our questions actually fall, so the time goes where the marks are likely to be. It is our reading of the syllabus, not the exam body's weighting.

  1. First, the heavy units (40% of the 84 questions): Cramer-Rao inequality and Fisher information (9, 11%); Maximum likelihood: properties, invariance and asymptotic normality (9, 11%); Iterative maximum likelihood: Newton-Raphson and Fisher scoring (8, 10%); Method of moments (8, 10%).

    Read the manual chapters for these first and do every worked question you can find on them.

  2. Then the middle band: Minimum chi-square and other methods (8, 10%); Minimum variance unbiased estimation; Rao-Blackwell and Lehmann-Scheffe theorems (8, 10%); Sufficiency, the factorisation theorem and minimal sufficiency (8, 10%); Interval estimation: pivotal quantities and coverage (7, 8%).

    Aim for one sitting per unit, then mixed questions across them.

  3. Then the lighter units: Point estimation: unbiasedness, consistency and efficiency (7, 8%); Completeness, ancillarity and Basu's theorem (6, 7%); Large-sample confidence intervals (6, 7%).

    Learn the definitions, classifications and standard numbers; these units are more often recalled than worked.

  4. Leave time for the hard ones: Cramer-Rao inequality and Fisher information (3 hard of 9), Completeness, ancillarity and Basu's theorem (2 hard of 6), Maximum likelihood: properties, invariance and asymptotic normality (2 hard of 9).

    Come back to these after a first pass, untimed, and read every explanation, including for the options you rejected.

  5. Bank the easy marks: Minimum chi-square and other methods (5 easy of 8); Point estimation: unbiasedness, consistency and efficiency (4 easy of 7).

    Do these early in revision so they are fast on the day.

  6. Finish timed. Sit the subject as a timed set without notes, mark it, and send each wrong answer back to its unit above.

Last updated · How Laisha writes and checks its pages · Report a mistake

Ask Laisha about this

Opens Laisha's chat with your question ready to send. Sign in or start free.

Unlock every question — IIT JAM Statistical Inference I: Estimation

US$10 a month · one price worldwide · cancel any time

  • Every practice question in this subject, with full working
  • Timed exam papers and progress by unit
  • Ask Laisha to explain any answer — up to 1,000 questions a month
Free or Smarter
FreeSmarter
Grow guides, pest pages, articlesRead allRead all
Ask Laisha5 a dayUp to 1,000 a month
Exam practiceThe sample on each pageEvery question, timed papers
Crop calendar and pest alerts—For your own fields
Calculators and tools—All
Uploads and reports—100 a month, real files
10 days free for every new account, no card.