Research & Science

What research and science employers assess

Research selection tends to combine numerical and data reasoning with inductive and critical thinking, which mirrors the work: interpreting a dataset, forming a hypothesis from a pattern, and judging whether evidence actually supports a claim. The sections are usually less about subject knowledge than about method, which is why they are trainable even if your degree is some years behind you.

The most common failure is statistical rather than arithmetic: reading a change in a small sample as a real effect, or treating correlation as a cause. Practising that distinction deliberately is worth more than another round of calculation drills.

Assessment types in research and science hiring

The mix depends on whether the role is laboratory-based, computational or analytical. All three commonly include a data interpretation section, which is where the largest improvement is usually available.

• Data interpretation. Tables, charts and small datasets with questions requiring a specific calculation or comparison. Often the largest section.
• Numerical reasoning. Rates, ratios, means and weighted averages, usually presented in a scientific or commercial context.
• Inductive and pattern reasoning. Sequences and rule-finding, measuring hypothesis formation rather than prior knowledge.
• Critical thinking. Judging whether a study design supports its conclusion, and identifying the assumption a claim depends on.
• Technical interview or journal discussion. Discussion of your own work, with attention to what the evidence does and does not show.

A method for data interpretation

Write the required quantity as a formula before reading the table, then check the sample size and the base before drawing any conclusion. Most marks are lost to the second step rather than the first.

1. Read the question and write down the quantity required, including any subgroup restriction.
2. Identify from the table only the rows or columns that feed that quantity.
3. For an average, check whether the groups being combined have equal sizes. If not, weight them.
4. For a change over time, check the base period and whether the chart axis starts at zero.
5. Before calling a difference real, check the sample size. Small samples produce large apparent differences by chance.
6. For a claim about cause, ask what else changed at the same time, and whether the comparison group is genuinely comparable.

Worked examples

Two laboratories report mean yields. Lab A reports a mean of 40 from 5 runs. Lab B reports a mean of 50 from 15 runs. What is the overall mean yield across all 20 runs?

Answer: 47.5

Why: An unweighted average of 40 and 50 gives 45 and is wrong because the groups differ in size. The totals are (40 x 5) = 200 and (50 x 15) = 750, giving 950 across 20 runs. 950 / 20 = 47.5. Checking group sizes before averaging is the single most useful habit in data interpretation sections.

A reaction takes 80 seconds at 20 degrees and 40 seconds at 30 degrees. For this reaction the time halves for every 10-degree rise in temperature. How long will it take at 40 degrees?

Answer: 20 seconds. The prompt states that the time halves for each 10-degree rise: 80 seconds at 20 degrees, 40 at 30, so 20 at 40. Naming the model you are applying is what makes the answer determinate.

Why: Two data points show the time halving across a 10-degree rise. Halving again from 40 gives 20 seconds. The important discipline is naming the assumption you are applying: with only two points the doubling is a pattern suggested by the data, not something the data proves. Critical thinking sections frequently test exactly this distinction.

A study finds that sites with more monitoring equipment report more incidents. Does the equipment cause incidents?

Answer: No, and the data cannot settle the question as presented. A more plausible reading is that better monitoring detects and records more of what was already happening.

Why: Both variables could be driven by a third factor such as site size or budget, and detection itself changes the count. The correct response is to identify the alternative explanation rather than to accept or reject the correlation outright. Questions like this appear in critical thinking and study-design sections.

Common mistakes

• Averaging groups of different sizes. Use a weighted mean whenever the group sizes differ.
• Reading a small difference as an effect. Check the sample size before drawing a conclusion from a gap.
• Treating correlation as cause. Ask what else changed and whether the comparison group is comparable.
• Estimating from a chart with a truncated axis. Compute from the printed values wherever they are given.
• Answering beyond the data. The strongest answer is usually the one the evidence supports, not the one that is most interesting.

Where to go next

Practise data interpretation with the aim of eliminating selection errors rather than increasing speed, and read every explanation for the items you answer correctly. In critical thinking practice, write down the assumption each argument depends on as a sentence you could test.

• Numerical reasoning practice
• Critical thinking practice
• Inductive reasoning practice
• Deductive reasoning practice
• All cognitive test intros

Frequently asked questions