LAISHA
Open Laisha

Indian Statistical Service Examination Statistical Inference I: Estimation: practice questions

Indian Statistical Service Examination (with Indian Economic Service) · India · Union Public Service Commission

Original practice questions written to the published syllabus · not past papers
10worked here, free
24written to this exam in this subject
11syllabus units

About Indian Statistical Service Examination

Indian Statistical Service Examination (with Indian Economic Service) is a government recruitment exam run by Union Public Service Commission, for candidates in India. Written examination of six papers and a viva voce. Reported scheme: General English 100, General Studies 100, Statistics I and II (objective, 200 each), Statistics III and IV (descriptive, 200 each); viva 200.

Official site: upsc.gov.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 Service Examination

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 24 Indian Statistical Service Examination questions in this subject: 7 easy, 9 medium, 8 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 Service Examination and are not shown on any other exam's page.

Question 1 · easy

Which estimator of σ² from a normal sample is consistent but biased?

  1. Σ(Xᵢ − X̄)²/(n − 1)
  2. (X₁ − X̄)²
  3. Σ(Xᵢ − X̄)²/n
  4. (X₁ − X₂)²/2
Answer and working

C. Σ(Xᵢ − X̄)²/n

The divisor-n estimator has expectation (n − 1)σ²/n → σ² and variance → 0, so it is consistent but biased. s² is unbiased and consistent; (X₁ − X₂)²/2 is unbiased but not consistent, since it uses only two observations.

Question 2 · easy

For a random sample from a population with mean μ, why is T = X₁ (the first observation) unbiased but not consistent?

  1. It is consistent, because every unbiased estimator is consistent in large samples
  2. Its variance falls like 1/n but its bias stays fixed at a non-zero value
  3. Its sampling distribution never narrows as n grows
  4. It is biased for every n, and the bias does not shrink as n grows
Answer and working

C. Its sampling distribution never narrows as n grows

E(X₁) = μ, but X₁ has the population distribution for every n, so P(|X₁ − μ| > ε) does not tend to 0. Consistency needs the sampling distribution to collapse onto μ.

Question 3 · easy

For a random sample from N(μ, σ²) with both parameters unknown, which statistic is jointly sufficient?

  1. (ΣXᵢ, ΣXᵢ²)
  2. (X₍₁₎, X₍ₙ₎)
  3. ΣXᵢ² alone
  4. ΣXᵢ alone
Answer and working

A. (ΣXᵢ, ΣXᵢ²)

The likelihood ∝ σ⁻ⁿ exp[−(ΣXᵢ² − 2μΣXᵢ + nμ²)/(2σ²)] depends on the data only through ΣXᵢ and ΣXᵢ², equivalently (X̄, s²). ΣXᵢ alone is sufficient only when σ² is known.

Question 4 · hard

For X₁, ..., Xₙ iid uniform on (θ, θ + 1), why is the minimal sufficient statistic (X₍₁₎, X₍ₙ₎) not complete?

  1. The range X₍ₙ₎ − X₍₁₎ has a θ-free distribution, so its centred version has mean 0 for every θ
  2. Its distribution depends on θ, which a complete statistic is not allowed to do
  3. It is two-dimensional, and a complete statistic must have exactly one dimension per parameter
  4. It is not unbiased for θ, and only unbiased statistics can be complete
Answer and working

A. The range X₍ₙ₎ − X₍₁₎ has a θ-free distribution, so its centred version has mean 0 for every θ

g = (range) − E(range) is a non-zero function of the statistic with zero mean for every θ, violating completeness. Dimension is not the issue; complete statistics can have any dimension, and completeness has nothing to do with unbiasedness.

Question 5 · easy

X₁, ..., Xₙ are iid N(θ, 1). Which statistic is ancillary for θ?

  1. The sample maximum X₍ₙ₎
  2. The sum ΣXᵢ
  3. The range X₍ₙ₎ − X₍₁₎
  4. The sample mean X̄
Answer and working

C. The range X₍ₙ₎ − X₍₁₎

In a location family, differences between observations do not depend on θ, so the range (and s²) are ancillary. X̄, the maximum and the sum all shift with θ.

Question 6 · medium

For a normal sample with both parameters unknown, s² is the UMVUE of σ². Which statement about estimating σ is correct?

  1. s is the UMVUE of σ, because a square root preserves unbiasedness
  2. No unbiased estimator of σ exists for normal samples of any size
  3. s is unbiased for σ, but it is not a function of the minimal sufficient statistic
  4. s is biased for σ; the UMVUE is c_n s with a gamma-function constant c_n > 1
Answer and working

D. s is biased for σ; the UMVUE is c_n s with a gamma-function constant c_n > 1

E(s) < σ by Jensen's inequality, so s is biased; multiplying by c_n = √((n − 1)/2)Γ((n − 1)/2)/Γ(n/2) makes it unbiased, and it is a function of the complete sufficient statistic, hence the UMVUE. Unbiasedness is not preserved by non-linear functions.

Question 7 · medium

X is a single binomial(n, p) observation with n known. Which parameter has NO unbiased estimator?

  1. p
  2. p²
  3. p(1 − p)
  4. 1/p
Answer and working

D. 1/p

E[g(X)] = Σ g(x)C(n, x)pˣ(1 − p)ⁿ⁻ˣ is a polynomial in p of degree at most n, so only polynomials of degree ≤ n can be estimated unbiasedly. 1/p is not a polynomial and is unbounded as p → 0, so no unbiased estimator exists; p, p(1 − p) and p² (for n ≥ 2) all have one.

Question 8 · medium

Five observations from uniform (0, θ) have mean 3.1 and maximum 6.5. What is the method-of-moments estimate, and what is wrong with it?

  1. 3.1, which is too small because it ignores the factor 2 in E(X) = θ/2
  2. 6.2, which is impossible because it is below the observed maximum of 6.5
  3. 7.8, which is fine but biased upward by the factor (n + 1)/n
  4. 6.5, which is fine because the moment estimate equals the maximum
Answer and working

B. 6.2, which is impossible because it is below the observed maximum of 6.5

E(X) = θ/2 gives θ̂ = 2X̄ = 6.2, yet θ must be at least 6.5. Moment estimates can fall outside the parameter space; the MLE X₍ₙ₎ = 6.5 or the UMVUE 7.8 respect it.

Question 9 · easy

Under absolute-error loss |θ − a|, the Bayes estimator of θ is:

  1. The posterior mode
  2. The posterior median
  3. The posterior mean
  4. The maximum-likelihood estimate
Answer and working

B. The posterior median

Expected absolute loss is minimised by the median of the posterior. Squared-error loss gives the posterior mean and 0-1 loss the posterior mode.

Question 10 · easy

A 95% confidence interval for mean yield is 3.8 to 4.6 t/ha. Which interpretation is correct?

  1. If the experiment were repeated, 95% of sample means would fall between 3.8 and 4.6
  2. 95% of individual plots will yield between 3.8 and 4.6 t/ha
  3. 95% of intervals built this way cover the true mean; this one either does or does not
  4. There is a 95% probability that the true mean lies between 3.8 and 4.6 t/ha for this trial
Answer and working

C. 95% of intervals built this way cover the true mean; this one either does or does not

The confidence level is a property of the method over repeated sampling. The probability statement about this fixed interval is a Bayesian credible-interval interpretation; the interval is for the mean, not for individual plots, and it does not predict future sample means.

A study plan for Statistical Inference I: Estimation in Indian Statistical Service Examination

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 (46% of the 24 questions): Completeness, ancillarity and Basu's theorem (3, 13%); Minimum variance unbiased estimation; Rao-Blackwell and Lehmann-Scheffe theorems (3, 13%); Point estimation: unbiasedness, consistency and efficiency (3, 13%); Cramer-Rao inequality and Fisher information (2, 8%).

    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 (2, 8%); Iterative maximum likelihood: Newton-Raphson and Fisher scoring (2, 8%); Large-sample confidence intervals (2, 8%); Method of moments (2, 8%).

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

  3. Then the lighter units: Minimum chi-square and other methods (2, 8%); Sufficiency, the factorisation theorem and minimal sufficiency (2, 8%); Maximum likelihood: properties, invariance and asymptotic normality (1, 4%).

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

  4. Leave time for the hard ones: Maximum likelihood: properties, invariance and asymptotic normality (1 hard of 1), Completeness, ancillarity and Basu's theorem (2 hard of 3), Iterative maximum likelihood: Newton-Raphson and Fisher scoring (1 hard of 2).

    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 (2 easy of 3); Interval estimation: pivotal quantities and coverage (1 easy of 2).

    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 Service Examination 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.