LAISHA
Open Laisha

ICAR AICE-JRF/SRF Bayesian Inference: practice questions

ICAR AICE-JRF/SRF - Agricultural Statistics, Computer Application, Bioinformatics (PhD admission and fellowship) · India · National Testing Agency for ICAR

Original practice questions written to the published syllabus · not past papers
10worked here, free
67written to this exam in this subject
11syllabus units

About ICAR AICE-JRF/SRF

ICAR AICE-JRF/SRF - Agricultural Statistics, Computer Application, Bioinformatics (PhD admission and fellowship) is a PhD, fellowship and lectureship exam run by National Testing Agency for ICAR, for candidates in India. Computer-based test, 2 hours, 120 MCQs, +4 / -1 (Part A general, Part B core group, Part C subject). 2026 exam held with AIEEA-PG in July 2026.

Source: icar.org.in.

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

Bayesian Inference in ICAR AICE-JRF/SRF

Inference through prior and posterior: conjugate analysis, prior choice, credible intervals, Bayes factors and decision theory, hierarchical models, and computation by Markov-chain Monte Carlo, with agricultural uses in variety trials, genomic prediction and small-area estimation.

Our 67 ICAR AICE-JRF/SRF questions in this subject: 8 easy, 32 medium, 27 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 ICAR AICE-JRF/SRF and are not shown on any other exam's page.

Question 1 · medium

A variety's mean yield has prior N(4.0, 0.5²) t/ha, and nine plots with known σ = 0.6 are observed. What is the posterior standard deviation of μ?

  1. 0.539 t/ha
  2. 0.186 t/ha
  3. 0.0345 t/ha
  4. 0.200 t/ha
Answer and working

B. 0.186 t/ha

Posterior precision = 4 + 25 = 29, variance 1/29 = 0.0345, SD 0.186. 0.200 is the data-only standard error; adding variances instead of precisions gives 0.539; 0.0345 is the variance.

Question 2 · medium

Plants are classified into three disease grades with counts 10, 5 and 5, and a Dirichlet(1, 1, 1) prior is used for the grade probabilities. What is the posterior mean probability of the first grade?

  1. 0.4783
  2. 0.3333
  3. 0.5000
  4. 0.5238
Answer and working

A. 0.4783

The posterior is Dirichlet(11, 6, 6); the mean of the first component is 11/(11 + 6 + 6) = 11/23. 0.5 is the MLE; 1/3 is the prior mean.

Question 3 · easy

How are posterior odds related to prior odds and the Bayes factor B₁₀?

  1. Posterior odds = B₁₀/prior odds
  2. Posterior odds = B₁₀, regardless of the prior
  3. Posterior odds = prior odds × B₁₀
  4. Posterior odds = prior odds + B₁₀
Answer and working

C. Posterior odds = prior odds × B₁₀

P(H₁ | x)/P(H₀ | x) = [P(H₁)/P(H₀)] × [m₁(x)/m₀(x)]. The Bayes factor is the factor by which the data change the odds.

Question 4 · medium

Two models differ in BIC by 6, with model 1 having the smaller BIC. Using 2 log B₁₀ ≈ BIC₀ − BIC₁, what is the approximate Bayes factor in favour of model 1?

  1. 3.0
  2. 20.1
  3. 6.0
  4. 403
Answer and working

B. 20.1

log B₁₀ ≈ 6/2 = 3, so B₁₀ ≈ e³ = 20.1. Forgetting the factor 2 gives e⁶ ≈ 403.

Question 5 · hard

What is Lindley's paradox?

  1. Improper priors give Bayes factors equal to 1 for every data set
  2. The posterior probability of H₀ can exceed 1 under diffuse priors
  3. With small n, the Bayes factor always agrees with the p-value
  4. With large n, a result significant at p = 0.01 can give a Bayes factor favouring the point null
Answer and working

D. With large n, a result significant at p = 0.01 can give a Bayes factor favouring the point null

For fixed p and growing n, the data concentrate near θ₀ relative to a diffuse alternative prior, so B₀₁ grows; frequentist and Bayesian conclusions can conflict. Posterior probabilities never exceed 1.

Question 6 · hard

A hierarchical model has only four trials. Why is a full Bayesian analysis with a prior on τ preferred to plugging in an estimate of τ?

  1. It propagates uncertainty in τ to the group posteriors
  2. It removes the need for any prior on the group effects
  3. It makes the group estimates identical to the raw means
  4. It always gives a larger estimate of τ than REML
Answer and working

A. It propagates uncertainty in τ to the group posteriors

With few groups τ is poorly determined; plug-in methods treat it as known and understate posterior uncertainty, while full Bayes integrates over it. Neither direction of the τ estimate is guaranteed.

Question 7 · hard

Viewed as a Metropolis-Hastings algorithm, what is the acceptance probability of a Gibbs update?

  1. 0.234
  2. The ratio of full conditionals
  3. 1
  4. 0.5
Answer and working

C. 1

With the full conditional as proposal, the MH ratio equals 1, so every draw is accepted.

Question 8 · hard

Which property of the Metropolis-Hastings transition ensures that the posterior is its stationary distribution?

  1. Conjugacy of the prior
  2. Detailed balance
  3. Symmetry of the proposal density
  4. Independence of successive draws
Answer and working

B. Detailed balance

π(θ)P(θ → θ*) = π(θ*)P(θ* → θ) holds by construction of α, which implies stationarity. Proposals need not be symmetric, and successive draws are dependent.

Question 9 · medium

What R̂ threshold do recent recommendations (Vehtari and colleagues, 2021) use for rank-normalised R̂?

  1. 1.1
  2. 2.0
  3. 1.5
  4. 1.01
Answer and working

D. 1.01

The older rule of thumb was R̂ < 1.1; the rank-normalised R̂ proposal recommends 1.01. Thresholds are conventions, and both are quoted in the literature.

Question 10 · hard

Which criterion is asymptotically equivalent to leave-one-out cross-validation and uses the whole posterior?

  1. WAIC
  2. Adjusted R²
  3. DIC
  4. BIC
Answer and working

A. WAIC

WAIC is built from the pointwise posterior predictive density and its variance and is asymptotically equivalent to LOO-CV. DIC uses a point estimate; BIC approximates the marginal likelihood.

A study plan for Bayesian Inference in ICAR AICE-JRF/SRF

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 (43% of the 67 questions): Empirical Bayes (8, 12%); Conjugate families (7, 10%); Markov chain Monte Carlo: Gibbs sampling and Metropolis-Hastings (7, 10%); Non-informative and weakly informative priors (7, 10%).

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

  2. Then the middle band: Posterior predictive checks and model comparison (DIC, WAIC) (7, 10%); Bayesian hypothesis testing and Bayes factors (6, 9%); Convergence diagnostics (6, 9%); Hierarchical (multilevel) models (6, 9%).

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

  3. Then the lighter units: Bayes' theorem for parameters: prior, likelihood and posterior (5, 7%); Decision theory: loss functions and Bayes risk (5, 7%); Posterior summaries and credible intervals (3, 4%).

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

  4. Leave time for the hard ones: Non-informative and weakly informative priors (5 hard of 7), Markov chain Monte Carlo: Gibbs sampling and Metropolis-Hastings (4 hard of 7), Empirical Bayes (4 hard of 8).

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

  5. Bank the easy marks: Bayesian hypothesis testing and Bayes factors (2 easy of 6); Bayes' theorem for parameters: prior, likelihood and posterior (1 easy of 5).

    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 — ICAR AICE-JRF/SRF Bayesian Inference

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.