LAISHA
Open Laisha

CSIR-UGC NET Categorical Data Analysis & Generalized Linear Models: practice questions

CSIR-UGC NET - Mathematical Sciences (statistics option) · India · National Testing Agency for CSIR

Original practice questions written to the published syllabus · not past papers
10worked here, free
42written to this exam in this subject
11syllabus units

About CSIR-UGC NET

CSIR-UGC NET - Mathematical Sciences (statistics option) is a PhD, fellowship and lectureship exam run by National Testing Agency for CSIR, for candidates in India. Computer-based, 3 hours, Parts A (general aptitude), B and C (subject); statisticians answer from the statistics unit.

Official site: csirnet.nta.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.

Categorical Data Analysis & Generalized Linear Models in CSIR-UGC NET

Models for counts, proportions and categories: contingency tables, the exponential-family GLM with its link functions, logistic and Poisson regression, log-linear models, overdispersion, ordinal outcomes and generalised linear mixed models, applied to germination, mortality, pest counts and adoption data.

Our 42 CSIR-UGC NET questions in this subject: 1 easy, 17 medium, 24 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 CSIR-UGC NET and are not shown on any other exam's page.

Question 1 · easy

What is the variance function V(μ) of the Poisson family?

  1. μ(1 − μ)
  2. 1
  3. μ
  4. μ²
Answer and working

C. μ

Poisson Var(y) = μ. The normal has V = 1, the gamma μ², the binomial proportion μ(1 − μ).

Question 2 · hard

For a density exp{[yθ − b(θ)]/a(φ) + c(y, φ)}, what are E(y) and Var(y)?

  1. b′(θ) and b″(θ)a(φ)
  2. b″(θ) and b′(θ)a(φ)
  3. θ and a(φ)
  4. b(θ) and b′(θ)a(φ)
Answer and working

A. b′(θ) and b″(θ)a(φ)

Differentiating the normalising identity gives E(y) = b′(θ) and Var(y) = b″(θ)a(φ) = V(μ)a(φ).

Question 3 · hard

With the canonical link, what form do the likelihood equations of a GLM take?

  1. (y − μ)′(y − μ) = 0
  2. X′(y − μ) = 0
  3. X′y = 0
  4. X′WX = I
Answer and working

B. X′(y − μ) = 0

With θ = η = Xβ, the score is X′(y − μ)/a(φ), so the equations are X′(y − μ) = 0 and X′y is sufficient for β.

Question 4 · hard

In IRLS for a Poisson log-linear model, what are the working weight and working response for observation i?

  1. wᵢ = μᵢ(1 − μᵢ) and zᵢ = yᵢ
  2. wᵢ = 1 and zᵢ = log yᵢ
  3. wᵢ = μᵢ and zᵢ = ηᵢ + (yᵢ − μᵢ)/μᵢ
  4. wᵢ = 1/μᵢ and zᵢ = ηᵢ + (yᵢ − μᵢ)μᵢ
Answer and working

C. wᵢ = μᵢ and zᵢ = ηᵢ + (yᵢ − μᵢ)/μᵢ

With g(μ) = log μ, g′(μ) = 1/μ and V(μ) = μ, so wᵢ = 1/[μᵢ(1/μᵢ)²] = μᵢ and zᵢ = ηᵢ + (yᵢ − μᵢ)/μᵢ. μ(1 − μ) is the binomial variance function.

Question 5 · medium

Trap A catches 42 thrips in 6 days and trap B 50 in 10 days. What is the standard error of the log rate ratio?

  1. 0.516
  2. 0.209
  3. 0.104
  4. 0.105
Answer and working

B. 0.209

For Poisson counts, Var(log count) ≈ 1/count, so SE = √(1/42 + 1/50) = 0.209; z = 0.336/0.209 = 1.61. Using the exposure times instead of counts is a slip.

Question 6 · hard

Why do Poisson and multinomial sampling give the same estimates of association in log-linear models?

  1. The multinomial is the limit of the Poisson for large n
  2. Both distributions have variance equal to the mean
  3. The total carries no association information
  4. Log-linear models ignore the sampling scheme entirely
Answer and working

C. The total carries no association information

Given the total, independent Poisson counts are multinomial; the total carries no information about association, so the kernels for the association parameters coincide (provided the model includes the terms fixed by design).

Question 7 · medium

How is the dispersion parameter usually estimated in a quasi-likelihood GLM?

  1. Residual deviance divided by n
  2. The largest Pearson residual squared
  3. The variance of the raw responses
  4. Pearson X² divided by the residual df
Answer and working

D. Pearson X² divided by the residual df

φ̂ = X²/(n − p) is the standard moment estimator; deviance/df is an alternative that can be biased for small counts.

Question 8 · medium

What is the deviance residual for a Poisson observation y = 7 with fitted mean 4?

  1. 3.00
  2. 1.50
  3. 1.83
  4. 1.35
Answer and working

D. 1.35

sign(y − μ̂)√{2[y log(y/μ̂) − (y − μ̂)]} = √1.83 = 1.35. 1.50 is the Pearson residual; 1.83 is the unit deviance.

Question 9 · hard

Why are randomised quantile residuals recommended for binary and small-count GLMs?

  1. Close to N(0, 1) under the true model
  2. They equal Pearson residuals for every family (and link)
  3. They are exactly zero for well-fitting observations
  4. They remove the need to choose a link function
Answer and working

A. Close to N(0, 1) under the true model

Pearson and deviance residuals fall in bands for discrete data; Dunn-Smyth residuals add randomisation within each discrete step so that they behave like standard normal deviates when the model is right.

Question 10 · medium

For a Poisson GLM with the canonical log link, which statistic is sufficient for β?

  1. X′X
  2. y′y
  3. Σ log yᵢ
  4. X′y
Answer and working

D. X′y

With θ = η = Xβ the log-likelihood is β′X′y − Σexp(xᵢ′β) − Σlog yᵢ!, so X′y is sufficient.

A study plan for Categorical Data Analysis & Generalized Linear Models in CSIR-UGC NET

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 (48% of the 42 questions): Exponential family and the generalized linear model (6, 14%); Generalized linear mixed models (5, 12%); Log-linear models for multi-way tables (5, 12%); Link functions, deviance and iteratively reweighted least squares (4, 10%).

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

  2. Then the middle band: Ordinal and multinomial logistic models (4, 10%); Overdispersion: quasi-likelihood and negative binomial models (4, 10%); Poisson regression for counts (4, 10%); Model checking: residuals and goodness of fit (3, 7%).

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

  3. Then the lighter units: Zero-inflated and hurdle models (3, 7%); Contingency tables: association, odds ratios and tests (2, 5%); Logistic regression for binary and binomial data (2, 5%).

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

  4. Leave time for the hard ones: Logistic regression for binary and binomial data (2 hard of 2), Generalized linear mixed models (4 hard of 5), Model checking: residuals and goodness of fit (2 hard of 3).

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

  5. Bank the easy marks: Exponential family and the generalized linear model (1 easy of 6).

    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 — CSIR-UGC NET Categorical Data Analysis & Generalized Linear Models

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.