LAISHA
Open Laisha

Indian Statistical Institute admission tests Statistical Inference I: Estimation: practice questions

Indian Statistical Institute admission tests (BStat, MStat, JRF in Statistics) · India · Indian Statistical Institute

Original practice questions written to the published syllabus · not past papers
10worked here, free
65written to this exam in this subject
11syllabus units

About Indian Statistical Institute admission tests

Indian Statistical Institute admission tests (BStat, MStat, JRF in Statistics) is a postgraduate admission exam run by Indian Statistical Institute, for candidates in India. Written tests in two parts: an objective mathematics/statistics test and a descriptive subject test, followed by interview for some programmes (reported). JRF selection by written test and interview.

Official site: isical.ac.in.

We have not been able to confirm these details against the exam body's own specification, so fees, dates and pass marks are left out on purpose. Check the current format, eligibility, fees and dates on the official site.

Statistical Inference I: Estimation in Indian Statistical Institute admission tests

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 65 Indian Statistical Institute admission tests questions in this subject: 32 easy, 28 medium, 5 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 Indian Statistical Institute admission tests and are not shown on any other exam's page.

Question 1 · easy

A random sample of size 10 is drawn from a population with variance 4. What is the expected value of Σ(Xᵢ − X̄)²/n, the divisor-n sample variance?

  1. 4.44
  2. 3.6
  3. 36.0
  4. 4.0
Answer and working

B. 3.6

E[Σ(Xᵢ − X̄)²] = (n − 1)σ² = 36, so dividing by n = 10 gives 3.6: biased downward by the factor (n − 1)/n. Dividing by n − 1 gives the unbiased s² with expectation 4.0; 36 is the expected sum of squares itself.

Question 2 · easy

A statistic T is complete for a family {f_θ} when which condition holds?

  1. The conditional distribution of the data given T does not depend on θ
  2. E_θ[g(T)] = 0 for all θ implies P_θ(g(T) = 0) = 1 for all θ
  3. The distribution of T does not depend on θ at all
  4. Var_θ(T) attains the Cramér-Rao lower bound for all θ
Answer and working

B. E_θ[g(T)] = 0 for all θ implies P_θ(g(T) = 0) = 1 for all θ

Completeness rules out non-trivial unbiased estimators of zero based on T. The second option defines sufficiency, the third ancillarity, and the last efficiency.

Question 3 · easy

Four observations from uniform (0, θ) have maximum 9.0. What is the UMVUE estimate of θ?

  1. 9.00
  2. 18.00
  3. 7.20
  4. 11.25
Answer and working

D. 11.25

(n + 1)X₍ₙ₎/n = 5 × 9/4 = 11.25, unbiased and a function of the complete sufficient X₍ₙ₎. 9.00 is the MLE, biased downward; 7.20 corrects in the wrong direction.

Question 4 · easy

A sample from uniform (0, θ) is 2.3, 5.1, 4.4, 0.8, 3.9. What is the maximum-likelihood estimate of θ?

  1. 6.6
  2. 5.1
  3. 3.3
  4. 6.12
Answer and working

B. 5.1

L(θ) = θ⁻⁵ for θ ≥ 5.1 and 0 otherwise, decreasing in θ, so θ̂ = X₍₅₎ = 5.1. 6.6 = 2X̄ is the moment estimate; 6.12 = 6 × 5.1/5 is the UMVUE.

Question 5 · easy

A Poisson model is fitted by minimum chi-square to counts in 5 classes, with λ estimated from the same data. How many degrees of freedom does the goodness-of-fit statistic have?

  1. 5
  2. 3
  3. 4
  4. 2
Answer and working

B. 3

df = classes − 1 − parameters estimated = 5 − 1 − 1 = 3. 4 forgets the estimated parameter.

Question 6 · medium

A germination rate p has a uniform (beta(1, 1)) prior. Three of 10 seeds germinate. What is the Bayes estimate of p under squared-error loss?

  1. 0.4000
  2. 0.3333
  3. 0.2500
  4. 0.3000
Answer and working

B. 0.3333

The posterior is beta(1 + 3, 1 + 7) = beta(4, 8), with mean 4/12 = 0.3333. 0.3000 is the MLE (the posterior mode under a uniform prior); 0.25 and 0.40 mis-add the prior counts.

Question 7 · easy

In the linear model y = Xβ + ε, when do the least-squares and maximum-likelihood estimators of β coincide?

  1. Never, because maximum likelihood estimates σ² with divisor n
  2. Only when the design matrix X is square and invertible
  3. Whenever the errors have mean zero, whatever their distribution
  4. When errors are iid normal with constant variance
Answer and working

D. When errors are iid normal with constant variance

With normal, independent, equal-variance errors, the log-likelihood is a constant minus the residual sum of squares/(2σ²), so maximising it minimises the sum of squares. The divisor issue concerns σ², not β.

Question 8 · medium

In 20 seed lots tested, none carry a pathogen (p̂ = 0). What happens to the Wald interval p̂ ± 1.96√(p̂(1 − p̂)/n)?

  1. It extends below 0, which is corrected by truncating at 0
  2. It collapses to the single point 0, a clearly misleading statement of certainty
  3. It gives a sensible interval from 0 to 0.098, close to the exact Clopper-Pearson interval
  4. It is undefined, since the formula divides by p̂
Answer and working

B. It collapses to the single point 0, a clearly misleading statement of certainty

With p̂ = 0 the estimated SE is 0, so the interval is [0, 0]. The Wilson or Clopper-Pearson interval gives a sensible upper limit (for zero events in n, roughly 3/n for 95%, the 'rule of three').

Question 9 · medium

Ten Poisson counts total 30. Starting from λ₀ = 2, Fisher scoring with nI(λ) = n/λ gives λ₁ = 2 + (30/2 − 10)/(10/2). What is λ₁, and why?

  1. 2.5, halfway between the start and the MLE
  2. 3, the MLE X̄, reached in one step because the scoring step is exact here
  3. 2.6667, the same as the Newton-Raphson step, since the Poisson family is exponential
  4. 5, because the step is the score divided by λ₀
Answer and working

B. 3, the MLE X̄, reached in one step because the scoring step is exact here

λ₁ = 2 + 5/5 = 3 = X̄. With λ as the parameter (not the canonical log λ), scoring and Newton-Raphson differ: Newton gives 2.6667. On the canonical scale η = log λ they would coincide.

Question 10 · easy

Near the maximum, how fast does Newton-Raphson converge?

  1. Linearly: the error falls by the same constant factor at each successive step
  2. Sub-linearly: the error falls like 1/k after k steps
  3. In exactly one step, for every log-likelihood
  4. Quadratically: the error is of order the square of the previous error
Answer and working

D. Quadratically: the error is of order the square of the previous error

Newton-Raphson has quadratic convergence near a simple root of the score. Fisher scoring and EM converge linearly; one-step convergence happens only for a quadratic log-likelihood.

A study plan for Statistical Inference I: Estimation in Indian Statistical Institute admission tests

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 (45% of the 65 questions): Minimum chi-square and other methods (8, 12%); Iterative maximum likelihood: Newton-Raphson and Fisher scoring (7, 11%); Maximum likelihood: properties, invariance and asymptotic normality (7, 11%); Sufficiency, the factorisation theorem and minimal sufficiency (7, 11%).

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

  2. Then the middle band: Interval estimation: pivotal quantities and coverage (6, 9%); Method of moments (6, 9%); Point estimation: unbiasedness, consistency and efficiency (6, 9%); Cramer-Rao inequality and Fisher information (5, 8%).

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

  3. Then the lighter units: Large-sample confidence intervals (5, 8%); Completeness, ancillarity and Basu's theorem (4, 6%); Minimum variance unbiased estimation; Rao-Blackwell and Lehmann-Scheffe theorems (4, 6%).

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

  4. Leave time for the hard ones: Completeness, ancillarity and Basu's theorem (1 hard of 4), Cramer-Rao inequality and Fisher information (1 hard of 5), Point estimation: unbiasedness, consistency and efficiency (1 hard of 6).

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

  5. Bank the easy marks: Point estimation: unbiasedness, consistency and efficiency (4 easy of 6); Minimum chi-square and other methods (5 easy of 8).

    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 — Indian Statistical Institute admission tests 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.