Statistics Viva Questions: Practise Clear, Defensible Answers

What to expect from a statistics viva

There is no single international statistics viva format. A university assessment, research defence and professional interview may test different material and use different marking rules. Check your course guide or the relevant organisation’s official information for the format, topics and assessment criteria that apply to you. The practice stations below are not claims about a particular exam.

Whatever the setting, a strong spoken answer does more than name a test or repeat a formula. It identifies the question, states the assumptions, interprets the result in context and explains what the analysis cannot establish. You should also be ready to explain why you chose one method over a plausible alternative.

A useful answer sequence is:

1. Name the target. Say what is being estimated or tested, and identify the population or comparison.
2. State the method and its conditions. Mention the assumptions that matter for this particular analysis.
3. Interpret the result in the units and context given. Keep statistical evidence separate from practical importance.
4. Give a limitation or next check. Say what additional information could change the conclusion.

Practise aloud rather than only reading notes. Aim to give a direct first answer, then add the reasoning that supports it. If you do not know a detail, say what you would check rather than disguising uncertainty with a confident guess.

Set up a practice viva

Start with the material you are expected to know: lecture notes, problem sets, research methods and any stated learning outcomes. Turn each topic into a prompt that requires a decision or an explanation. “Define a confidence interval” is useful for recall; “Interpret this interval and tell me what it does not mean” is closer to a viva exchange.

In MySummaries, the examiner strip for this practice set looks like this:

Use the emphases as prompts for practice, not as official criteria. A lecturer or examination body may use a different rubric. For each station, rehearse a short first response, then prepare for follow-up questions such as “What assumption matters most here?” or “Would this establish causation?”

A useful practice set moves between inference, modelling, study design and interpretation. The stations below use common statistical ideas, but you should swap in topics from your own syllabus where needed. The table gives each station an opening prompt; the full practices that follow show how to develop an answer under questioning.

The station table for this set is:

Station 1: interpret an estimate and interval

This station tests whether you can describe an estimate without turning a confidence interval into a probability statement about a fixed parameter. The important distinction is between uncertainty in the estimation procedure and the particular interval calculated from this sample.

The examiner asks: “A study estimates a mean difference of 2.4 units, with a 95% confidence interval from 0.3 to 4.5. Interpret the result and explain what the interval means.” A focused answer gives the direction, size and uncertainty, then states what the interval does and does not imply.

A marked practice response might look like this:

That is a MySummaries station, filled with statistics material. Yours is written from your own notes. Start free

The key repair is to avoid saying, “There is a 95% probability that the true mean difference is between 0.3 and 4.5.” In the usual frequentist interpretation, the population parameter is treated as fixed; the interval-producing procedure has the repeated-sampling coverage property. Also avoid claiming that statistical significance proves a clinically or practically important effect. The interval’s width and the outcome’s units both matter.

If asked what you would check next, mention how the interval was calculated, whether the observations and sampling design support the method, and whether the difference is important on the relevant scale. Do not assume those details from the interval alone.

Station 2: interpret an adjusted regression

This station tests the boundary between an adjusted association and a causal claim. Adjustment for measured variables can change an estimate, but the word “adjusted” does not by itself remove confounding or establish causality.

The examiner asks: “In an observational study, a regression estimates an adjusted outcome difference of 1.8 units per additional study hour. What can you conclude, and what can you not conclude?” The strongest answer makes clear what the coefficient represents and asks what was adjusted for.

A marked practice response might look like this:

A common mark-losing phrase is “each extra hour causes an increase of 1.8”. The model reports a conditional association unless the design and causal assumptions justify more. The exact interpretation also depends on what the outcome is and how the model is specified; do not call a coefficient a mean difference if the model’s outcome or link function does not support that reading.

When practising this station, make your answer conditional: “If the outcome is measured in these units and the model is linear…” That signals that you understand a coefficient is not interpreted independently of the model.

Station 3: reason from prevalence and test accuracy

This station tests whether you can use sensitivity and specificity to answer a question about a person with a positive test. The base rate matters: the probability of disease after a positive result is not the same as sensitivity.

The examiner asks: “A test has 90% sensitivity and 95% specificity. If prevalence is 2%, estimate the probability that a person with a positive result has the condition.” Give the calculation in words or with Bayes’ rule, then explain why the result is lower than 90%.

A marked practice response might look like this:

A useful way to check your reasoning is to build the 1,000-person table aloud. The expected true positives are 0.90 × 20 = 18; expected false positives are 0.05 × 980 = 49. So the positive predictive value is 18 / (18 + 49) ≈ 0.269. Rounding to about 27% is appropriate.

The strongest distractor is “90%”: that is the sensitivity, or the probability of a positive result given the condition. The question reverses the condition and asks for the probability of the condition given a positive result. A viva examiner may follow up by asking what happens if prevalence rises. With sensitivity and specificity held constant, the positive predictive value rises as prevalence rises.

Review your spoken answers

After each attempt, review the answer for missing reasoning rather than just its score. Did you state what the number refers to? Did you give the assumptions that affect the conclusion? Did you mistake association for causation, or sensitivity for predictive value? Pick one correction and repeat the station with that correction included.

For example, a transcript can show that the conclusion was broadly sound but too brief about the calculation:

Treat the wording feedback as a prompt to practise a more precise statement. In particular, separate the probability calculation from the population assumptions: if prevalence or test performance differs in the population being tested, the predictive value can change.

You can make a short spoken debrief part of the review, then repeat the answer using the corrected wording:

How MySummaries helps

Build a statistics revision board from your own notes, then use it to generate oral stations, record spoken answers and review feedback against your study material. Adapt the prompts to the syllabus or assessment information that applies to you. Start at MySummaries.