Research Methods: Choosing a Design That Fits the Question

Research methods are choices about what can be learned from a question, with a particular population, under real constraints. An impressive technique cannot rescue a vague question or a sample that does not represent the people about whom the writer concludes. A sound design explains the question, the kind of evidence needed, how people or materials will be selected, how observations will be made, and which alternative explanations remain. The method section should let another reader judge the strength and limits of the resulting inference.

Begin with the claim the study must support

Distinguish a descriptive question from a causal or interpretive one. “How often do students use the library?” asks for a defined measure in a defined population and period. “Does extending library hours increase use?” requires a comparison and attention to other changes. “How do students experience access?” calls for a way to hear varied accounts in context. The same topic can support all three, but the questions demand different evidence. State the main outcome or phenomenon and the decision the result will inform.

Write the intended conclusion before choosing a method, then ask what would justify it. If the goal is to estimate a population percentage, a convenience sample of volunteers may be too narrow. If the goal is to understand a process, a well-chosen set of interviews may illuminate mechanisms without estimating prevalence. Match the claim to the design instead of claiming that one approach is universally stronger.

Define the population and recruitment path

Name who or what qualifies for the study, where they can be reached, and who the results are meant to describe. Explain inclusion and exclusion criteria with reasons. Sampling choices affect both feasibility and interpretation: a survey distributed in one classroom may miss remote students, while interviews recruited through a service may overrepresent people who already use it. Keep a record of invitations, participation, and nonresponse when appropriate, and explain what cannot be inferred about those absent.

Sample size should answer the study’s purpose. A quantitative design needs enough observations for a useful estimate or comparison, allowing for uncertainty and anticipated missing data. A qualitative design needs sufficient variation and depth to address the question, with a transparent account of how participants and cases were selected. A large sample with systematic exclusion can still mislead. Do not turn a numerical target into a claim of representativeness.

Measure the concept that matters

Translate broad terms into observable indicators without losing their meaning. If a study concerns learning, a satisfaction question is not a direct measure of learning. If it concerns access, login counts may omit people who never reach the service. Explain the instrument, timing, and unit of analysis. Consider whether an established measure fits the language and setting, whether questions lead respondents, and whether repeated measurements are comparable.

Pilot the procedure with a small group when practical. Ask whether instructions are understood, whether a response category is missing, and whether records can be linked without exposing private information. Record changes made after the pilot. Consistency matters, but a consistently measured proxy can still be invalid for the intended construct. Where several indicators capture different dimensions, explain what each adds rather than hiding them inside a single score.

Plan comparisons and analysis in advance

For a quantitative comparison, specify the groups, time window, outcome, likely confounders, and analysis before seeing results. A before-and-after change may reflect season, selection, or a concurrent policy rather than the intervention alone. A comparison group can improve inference when its circumstances are sufficiently similar; it does not automatically eliminate bias. Describe missing values and uncertainty instead of silently dropping difficult observations.

For qualitative work, define how records will be created, how themes or categories will be developed, and how contradictory cases will be handled. Keep the participant’s words distinct from the researcher’s interpretation. Reflexive notes can document how the researcher’s position affects questions and reading. Combining methods can be useful when one supplies a pattern and another explores its mechanisms, but each still needs a clear role and standard of quality.

Work through a campus service example

Suppose a university wants to know whether evening tutoring increases first-year course completion. Administrative records can compare completion before and after a schedule change, but students choosing evening sessions may differ in work hours and prior preparation. A stronger plan records eligibility, uptake, baseline preparation, and outcomes for a comparable period, then examines how selection and other campus changes could affect the estimate. Interviews with users and nonusers can reveal transport, timing, awareness, and perceived usefulness that the records miss.

The two strands answer different questions. A completion difference may be consistent with benefit without proving that tutoring caused it. Interviews may suggest why access changed but cannot estimate an effect for all students. The recommendation could be to continue a pilot, improve outreach, and measure a further term rather than claim definitive success. The design makes that limited decision possible because its assumptions and missing evidence are visible.

Protect participants and report limits honestly

Consent, privacy, data minimization, and the power relationship between researcher and participant matter from recruitment through publication. Explain how records will be stored, who can see them, and how small groups will be described without identification. Institutional review requirements vary; a classroom exercise does not automatically authorize the collection of sensitive information. Use the approval and supervision appropriate to the setting before gathering data.

In the final report, separate what was observed from what was inferred. State the scope, sample, instrument, analysis, unexpected deviations, and plausible alternative explanations. A study can be useful without answering every question. Its value lies in a conclusion whose strength matches the design and in a clear account of what another study or decision maker should do next.