LAISHA
Open Laisha

Indian Statistical Institute admission tests Categorical Data Analysis & Generalized Linear Models: 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
73written 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.

Categorical Data Analysis & Generalized Linear Models in Indian Statistical Institute admission tests

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 73 Indian Statistical Institute admission tests questions in this subject: 15 easy, 45 medium, 13 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

What is the canonical link for binomial proportions?

  1. Probit
  2. Logit
  3. Identity
  4. Log
Answer and working

B. Logit

The binomial natural parameter is log[p/(1 − p)]. The probit is a common non-canonical link.

Question 2 · medium

Why is a binomial GLM preferred to least-squares regression on arcsine-transformed germination percentages?

  1. It always gives smaller standard errors
  2. It works only when all percentages are between 20% and 80%
  3. It models the mean on its natural scale with the correct binomial variance and weights
  4. It needs no assumption about the variance
Answer and working

C. It models the mean on its natural scale with the correct binomial variance and weights

The GLM keeps fitted probabilities in (0, 1), weights by the number of seeds, and uses V(μ) = μ(1 − μ) directly; transformations only approximate this and struggle with 0% and 100%.

Question 3 · medium

For which binary-response situation is the complementary log-log link natural?

  1. Infection occurs if at least one of a Poisson number of propagules succeeds
  2. The tolerance distribution is normal
  3. The response is symmetric about p = 0.5
  4. The predictor is a percentage
Answer and working

A. Infection occurs if at least one of a Poisson number of propagules succeeds

If propagules arrive as Poisson(λ) with log λ = x′β, then P(infected) = 1 − exp(−λ), i.e. log[−log(1 − p)] = x′β. A normal tolerance gives the probit.

Question 4 · medium

What is the unit Poisson deviance contribution dᵢ for y = 7 with fitted mean 4?

  1. 3.92
  2. 1.35
  3. 2.25
  4. 1.83
Answer and working

D. 1.83

dᵢ = 2[y log(y/μ̂) − (y − μ̂)] = 2[7 log 1.75 − 3] = 1.83. 2.25 is the Pearson contribution (y − μ̂)²/μ̂; 1.35 is the deviance residual √dᵢ.

Question 5 · medium

A logistic regression fails to converge, with one coefficient and its standard error growing very large. What is the likely cause and remedy?

  1. Too few iterations: increase the iteration limit
  2. Overdispersion: use a quasi-binomial model
  3. Separation; use Firth's penalised likelihood
  4. A wrong link: switch from logit to probit
Answer and working

C. Separation; use Firth's penalised likelihood

Under complete or quasi-complete separation the MLE is infinite, so estimates diverge; Firth's bias-reduced likelihood or exact methods give finite estimates. More iterations or another link do not help.

Question 6 · hard

For a logit link, by roughly what factor do probit coefficients differ from logit coefficients fitted to the same data?

  1. They are identical, since both links are symmetric
  2. Probit coefficients are larger, about 1.6 to 1.8 times
  3. Probit coefficients are smaller, about 1/1.6 to 1/1.8 of the logit ones
  4. Probit coefficients are about π times the logit ones
Answer and working

C. Probit coefficients are smaller, about 1/1.6 to 1/1.8 of the logit ones

The logistic distribution has SD π/√3 ≈ 1.81 against 1 for the standard normal, so logit coefficients are larger by about 1.6-1.8. The exact ratio depends on the data.

Question 7 · medium

Why is AIC not available for a quasi-Poisson model?

  1. Its deviance is always zero
  2. AIC requires a normal response
  3. It has no full likelihood
  4. It has no free parameters
Answer and working

C. It has no full likelihood

Without a distribution there is no maximised log-likelihood; quasi-AIC (QAIC) variants exist but rest on approximations.

Question 8 · easy

In a hurdle model for pest counts, where do zeros come from?

  1. From neither; zeros are excluded
  2. Only from the count part
  3. Only from the binary part
  4. From both the binary and count parts
Answer and working

C. Only from the binary part

The count part is zero-truncated, so every zero comes from the binary zero-versus-positive model. In zero-inflated models zeros come from both sources.

Question 9 · medium

Market-sales quantities are zero for households that decide not to sell and positive for those that sell. Which model reflects this two-step decision?

  1. A hurdle model
  2. A cumulative logit model
  3. A log-linear model
  4. A zero-inflated Poisson
Answer and working

A. A hurdle model

A decision to sell (binary) followed by a quantity among sellers (truncated count or positive model) is the hurdle structure.

Question 10 · medium

For a 2 × 2 × 2 table, how many residual degrees of freedom does the conditional-independence model (XZ, YZ) have?

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

C. 2

Parameters: 1 + 3 main effects + 2 two-factor terms = 6; 8 − 6 = 2. Mutual independence has 4 and homogeneous association 1.

A study plan for Categorical Data Analysis & Generalized Linear Models 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 (42% of the 73 questions): Link functions, deviance and iteratively reweighted least squares (8, 11%); Ordinal and multinomial logistic models (8, 11%); Poisson regression for counts (8, 11%); Contingency tables: association, odds ratios and tests (7, 10%).

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

  2. Then the middle band: Model checking: residuals and goodness of fit (7, 10%); Zero-inflated and hurdle models (7, 10%); Exponential family and the generalized linear model (6, 8%); Generalized linear mixed models (6, 8%).

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

  3. Then the lighter units: Overdispersion: quasi-likelihood and negative binomial models (6, 8%); Log-linear models for multi-way tables (5, 7%); Logistic regression for binary and binomial data (5, 7%).

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

  4. Leave time for the hard ones: Exponential family and the generalized linear model (2 hard of 6), Generalized linear mixed models (2 hard of 6), Model checking: residuals and goodness of fit (2 hard of 7).

    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 (2 easy of 6); Overdispersion: quasi-likelihood and negative binomial models (2 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 — Indian Statistical Institute admission tests Categorical Data Analysis & Generalized Linea

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.