ASRB NET Linear Mixed Models, Variance Components & Spatial Analysis of Field Trials: practice questions
ASRB NET - Agricultural Statistics (and Computer Application / Bioinformatics) · India · Agricultural Scientists Recruitment Board (ICAR)
Original practice questions written to the published syllabus · not past papers
10worked here, free
84written to this exam in this subject
10syllabus units
About ASRB NET
ASRB NET - Agricultural Statistics (and Computer Application / Bioinformatics) is a professional certification run by Agricultural Scientists Recruitment Board (ICAR), for candidates in India. NET: one objective paper of about 150 questions, 2 hours, negative marking, online, Master's level (pattern as reported for other disciplines).
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.
Linear Mixed Models, Variance Components & Spatial Analysis of Field Trials in ASRB NET
Applied mixed-model analysis: REML versus ANOVA estimation, BLUP for selection, spatial analysis of variety trials, choosing covariance structures and software.
Models with more than one source of random variation: ANOVA, ML and REML estimation of variance components, BLUE and BLUP, Henderson's mixed-model equations, and spatial models that correct field trials for fertility trends and neighbour correlation.
Our 84 ASRB NET questions in this subject: 22 easy, 40 medium, 22 hard, across 10 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 ASRB NET and are not shown on any other exam's page.
Question 1 · easy
In the mixed model y = X beta + Z u + e, with u ~ N(0, G) and e ~ N(0, R), what is the variance of y?
G + R
Z R Z' + G
Z G Z' + R
X G X' + R
Answer and working
C. Z G Z' + R
V(y) = V(Zu) + V(e) = ZGZ' + R, since u and e are independent and X beta is fixed.
Question 2 · medium
Which model suits germination counts per pot with random blocks?
A linear mixed model on the raw counts, whatever their distribution
An ordinary least squares model with no blocks
A generalised linear mixed model with a suitable link
A fixed-effects ANOVA on block totals
Answer and working
C. A generalised linear mixed model with a suitable link
Counts or proportions with random effects are naturally handled by GLMMs (e.g. binomial or Poisson with a link function).
Question 3 · medium
In the mixed model, what is beta-hat from the mixed-model equations?
The BLUP of the random effects
The ML estimate of variance
The OLS estimate, ignoring the covariance matrix V
The BLUE of beta, equal to the GLS estimator
Answer and working
D. The BLUE of beta, equal to the GLS estimator
beta-hat solves the GLS problem with V = ZGZ' + R; u-hat is the BLUP.
Question 4 · easy
Why is spatial analysis used in field trials?
Fertility varies smoothly and irregularly, not only between blocks
Blocks are never useful in field trials
It removes the need for replication
Plots far apart are always more alike than adjacent plots in a block
Answer and working
A. Fertility varies smoothly and irregularly, not only between blocks
Neighbouring plots are more alike; spatial models capture trends that blocks miss.
Question 5 · medium
Can a spline surface and an AR1 x AR1 residual be combined in one model?
Only if the trial has a single row
Only when no genotypes or treatments are fitted in the model
No, they are mathematically incompatible
Yes, for large-scale trend plus local correlation
Answer and working
D. Yes, for large-scale trend plus local correlation
Combining them is possible; the cost is complexity and convergence risk.
Question 6 · hard
Why is the naive chi-square test conservative when testing a variance component equal to zero?
Variance components are always large
REML likelihoods cannot be compared between any two models at all
The null value lies on the boundary of the parameter space
The chi-square has too many df
Answer and working
C. The null value lies on the boundary of the parameter space
On the boundary the statistic follows a 50:50 mixture, so the naive p-value is twice the correct one.
Question 7 · medium
Comparing AR(1) and Toeplitz residual structures with the same fixed effects, which tool is appropriate?
Bartlett's test on residuals
REML-based AIC or BIC
A t test on the variety means
ML-based F tests on fixed effects
Answer and working
B. REML-based AIC or BIC
Covariance structures with common fixed effects are compared by REML likelihoods and information criteria.
Question 8 · medium
Why are replicate and block random terms usually kept even if not significant?
They always reduce AIC
They are required by all software
They reflect the randomisation of the design
They are fixed effects in disguise and are tested like treatments
Answer and working
C. They reflect the randomisation of the design
Design terms protect the validity of inference; significance testing is reserved for optional terms.
Question 9 · easy
What does REML stand for?
Repeated measures linear model
Restricted (residual) maximum likelihood
Random effects minimum loss
Regression estimate of mean level
Answer and working
B. Restricted (residual) maximum likelihood
REML maximises the likelihood of error contrasts.
Question 10 · easy
What does BLUP stand for?
Bayesian likelihood under priors
Best linear unbiased prediction
Balanced linear unit prediction
Block level uniform precision
Answer and working
B. Best linear unbiased prediction
BLUP predicts random effects; BLUE estimates fixed effects.
A study plan for Linear Mixed Models, Variance Components & Spatial Analysis of Field Trials in ASRB 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.
First, the heavy units (46% of the 84 questions): Henderson's mixed-model equations; BLUE and BLUP (11, 13%); Covariance structures for repeated measures (10, 12%); Maximum likelihood and REML (9, 11%); Spatial analysis of field trials: nearest-neighbour and AR1 x AR1 models (9, 11%).
Read the manual chapters for these first and do every worked question you can find on them.
Then the middle band: Fixed, random and mixed models revisited (8, 10%); Model selection with likelihood-ratio tests and information criteria (8, 10%); Unbalanced data and Satterthwaite and Kenward-Roger degrees of freedom (8, 10%); ANOVA (Henderson's methods) estimation of variance components (7, 8%).
Aim for one sitting per unit, then mixed questions across them.
Then the lighter units: Software: lme4, nlme, ASReml, SAS PROC MIXED (7, 8%); Trend and spline adjustment (e.g. SpATS approach) (7, 8%).
Learn the definitions, classifications and standard numbers; these units are more often recalled than worked.
Leave time for the hard ones: Model selection with likelihood-ratio tests and information criteria (3 hard of 8), Unbalanced data and Satterthwaite and Kenward-Roger degrees of freedom (3 hard of 8), Henderson's mixed-model equations; BLUE and BLUP (4 hard of 11).
Come back to these after a first pass, untimed, and read every explanation, including for the options you rejected.
Bank the easy marks: Unbalanced data and Satterthwaite and Kenward-Roger degrees of freedom (3 easy of 8); Covariance structures for repeated measures (3 easy of 10).
Do these early in revision so they are fast on the day.
Finish timed. Sit the subject as a timed set without notes, mark it, and send each wrong answer back to its unit above.