Hypotheses, Questions and Investigation Constraints

From Curiosity to Question

Every statistical investigation begins with a question someone wants to answer. A teacher wonders whether students who sleep longer perform better in tests. A public health team asks whether exercise frequency is linked to reported wellbeing. A market researcher wants to know whether younger shoppers spend more online than older ones. In each case, the curiosity is real — but curiosity alone is not enough. Statistics demands that you turn your question into something testable.

A hypothesis is a precise, testable statement — not a vague hunch. Compare these two versions of the same idea:

• Vague question: “Do students who sleep more do better?”
• Testable hypothesis: “Students who sleep eight or more hours per night will score higher on a memory test than students who sleep fewer than eight hours.”

The second version names the variables, states a direction, and makes clear what data you would need to collect. That precision matters because your entire investigation — what data to gather, how to sample, which statistical tools to use, and how to judge whether your results support or refute your prediction — flows from the hypothesis you write at the start.

Why Hypotheses Can Only Be Tested Through Data

It is tempting to think that a well-reasoned argument can “prove” a hypothesis. Statistics says no. You can believe something is true for entirely logical reasons and still be wrong when the data is collected. A hypothesis can only be tested through the appropriate collection and analysis of data — this is a core principle of the specification. Opinion, intuition, and personal experience are starting points for forming a hypothesis, not evidence for evaluating one.

This is why the design of an investigation matters so much. If your data collection is flawed — biased sample, poorly worded questions, or measurement error — then your conclusion, however confidently stated, rests on shaky foundations.

Investigation Constraints

Real investigations face practical limits. The specification lists five main types of constraint you should be able to discuss: time constraints, cost constraints, ethical issues, confidentiality concerns, and convenience.

Time: A survey conducted over one week gives a very different result from one run over a full year. If you are studying seasonal behaviour — ice cream sales, flu cases, holiday bookings — and your budget only allows one month of data collection, your results may not reflect the full picture.

Cost: A large random sample is statistically ideal but expensive. Interviewing participants in person is richer than an online questionnaire but far costlier. Many investigators compromise on sample size or method because of limited funds, and this affects the conclusions they can draw.

Ethical issues: Investigations involving human participants must not cause harm, distress, or embarrassment. Researching sensitive topics — mental health, financial hardship, relationship difficulties — requires careful handling. Participants must give their consent, and the researcher must decide whether the benefit of the knowledge outweighs any risk of harm.

Confidentiality: Data collected about individuals must be protected. If people fear that their personal information will be shared or used against them, they may refuse to participate or give dishonest answers. Guaranteeing anonymity — not attaching names to responses — typically improves both response rates and honesty.

Convenience: Sometimes the ideal population is simply not accessible. A researcher wanting to survey a random sample of all UK secondary school students faces enormous logistical challenges. Convenience shapes who can realistically be included.

Proactive Strategies: Planning for Problems

Good investigators do not just hope that problems will not arise — they plan for them. The specification expects you to know what kinds of problems can occur during an enquiry and how to mitigate them before they happen. Three common challenges are: difficulty identifying the population, non-response, and unexpected outcomes.

Identifying the population: For some investigations, the population is easy to define — all Year 11 students in a school, for example. For others, the boundary is blurry. If you want to study “regular gym users”, what counts as regular? Once per week? Five times per week? Defining the population clearly before you begin prevents disputes about who should have been included later.

Non-response: If a large proportion of people choose not to respond to your questionnaire, the results may be biased. Those who respond might systematically differ from those who do not. For example, people with strong opinions tend to respond at higher rates than those with moderate views. Strategies to reduce non-response include: keeping questionnaires short, offering an incentive, and following up with reminders.

Unexpected outcomes: Real data surprises you. A value that seems impossibly large might be a genuine outlier, a data entry error, or evidence that your hypothesis needs refinement. A good investigator builds in a review stage rather than assuming everything will go to plan.