Exam Strategy

MRCPCH TAS Statistics: 5 Mistakes That Cost Marks

Last reviewed: November 2026

Statistics questions in MRCPCH Theory and Science (TAS) rarely need more than simple arithmetic. They punish choosing the wrong measure, or answering a slightly different question from the one asked. Here are five common traps, each with a worked example and a fix.

Why naming the mistake matters

Statistics marks are rarely lost to difficult maths. They are usually lost to a small number of recognisable slips, such as using the right calculation for the wrong question. Once you can name the slip, a one-line habit usually prevents it.

Mistake 1: Using relative risk reduction when the question wants NNT

In a trial, 20% of the control group and 15% of the treatment group develop a complication.

  • Absolute risk reduction (ARR) = 20% − 15% = 5% (0.05)
  • Relative risk reduction (RRR) = 5 ÷ 20 = 25%
  • Number needed to treat (NNT) = 1 ÷ ARR = 1 ÷ 0.05 = 20

The trap is dividing 1 by the RRR (1 ÷ 0.25 = 4), which gives a wrong answer that is sure to be among the options. NNT always comes from the absolute difference, and if it is not a whole number you round up.

Fix: Write down the ARR first, every time.

Mistake 2: Mixing up sensitivity and positive predictive value

Sensitivity asks: of the children who have the condition, how many test positive? Positive predictive value (PPV) asks the reverse: of the children who test positive, how many have the condition? They sound alike, and PPV changes with how common the condition is.

Take a screening test with 95% sensitivity and 95% specificity, used in 1,000 children where 1 in 100 has the condition:

  • 10 children have it, and the test finds 9.5 of them (call it 10).
  • 990 do not, and 5% of them, 49.5, test positive anyway.
  • PPV = 9.5 ÷ (9.5 + 49.5), which is about 16%.

So most positive results are false positives, even though the test sounds excellent.

Fix: Draw a 2×2 table, fill it for 1,000 imaginary children, then read off the row or column the question asks about.

Mistake 3: Reading a confidence interval the wrong way round

A 95% confidence interval that includes the "no effect" value means the result is not statistically significant at the 5% level. The no-effect value depends on the measure: 1 for ratios (relative risk, odds ratio, hazard ratio) and 0 for differences (difference in means, absolute risk difference).

For example, a relative risk of 0.80 with a 95% confidence interval of 0.60 to 1.05 includes 1. The estimate suggests a 20% reduction, but the study has not shown a significant effect. A wider interval also means less precision, usually from a smaller sample.

Fix: Before reading the numbers, ask "ratio or difference?", then find the no-effect value.

Mistake 4: Treating a p-value as the chance the finding is true

A p-value of 0.03 does not mean there is a 97% chance the treatment works. It means that if there were truly no effect, results at least this extreme would turn up about 3 times in 100.

A very large study can also produce a p-value below 0.001 for a difference too small to matter to a child or family. Statistically significant is not the same as clinically important.

Fix: If an option says the p-value gives the probability that the hypothesis is true, it is almost always the distractor. Read the p-value together with the effect size and the confidence interval.

Mistake 5: Naming the wrong study design

Match the question to the design before you look at the options:

  • Does a treatment work? Randomised controlled trial.
  • What predicts an outcome, or what happens over time? Cohort study.
  • A rare condition, looking back for exposures? Case–control study.
  • How common is a condition right now? Cross-sectional study.
  • How well does a test perform? Diagnostic accuracy study.

Fix: Underline the verb in the question (treat, predict, estimate how common, diagnose) and map it to a design first.

A practical example

Consider a candidate who gets most statistics questions right in untimed practice but drops marks under time pressure. The likely cause is rarely missing knowledge. More often it is one of the habits above: starting the calculation before deciding which measure the question wants, or reading a confidence interval without first checking whether it is a ratio or a difference.

What to do this week

  1. Do five statistics questions. Before you read the options, write down the measure each one is asking for: ARR, NNT, PPV, confidence interval, p-value or study design.
  2. If you get one wrong, label the error as a knowledge gap, a missed clue or a misread task. Our post on keeping an error log that actually helps shows how.
  3. Read why you shouldn't revise FOP, TAS and AKP the same way to see why TAS needs its own approach.

Closing thought

Most TAS statistics errors are errors of selection, not calculation. Naming the measure first is often the quickest way to remove them.

Want more practice? Try the free 3-Minute MRCPCH and see how your measure-first habit holds up.

Original examples only, never recalled exam content. For postgraduate medical education and exam preparation. Always follow current guidance.