Sampling, data collection and research designs

Key ideas

Sampling

In probability sampling every member of a defined population has a known chance of selection, which supports generalizing to that population. Convenience and purposive samples are easier to obtain and suit many questions, but conclusions reach only as far as the people studied.

Experimental and observational designs

In experimental studies the investigator assigns the exposure, ideally at random. Observational studies only observe: cohort studies follow people forward from exposure, case-control studies trace back from outcome, and cross-sectional studies measure both at one time.

Descriptive designs

Case studies, case series and narrative research describe people or events in depth without a comparison group. They are valuable for generating hypotheses and understanding experience, but without a comparison they cannot test associations.

Quasi-experiments

A quasi-experiment tests an intervention without random assignment, for example by comparing intact groups or by measuring repeatedly before and after a change. Its strength depends on how well the design rules out rival explanations such as preexisting differences, history and maturation.

Single-case experimental designs

Repeated measurement of one person or a few people across phases, such as baseline and intervention, with control shown by replicating the effect within the study. Visual analysis of the graphed data remains the most common method of analysis.

Community-based participatory research

Community members, organizational representatives and researchers take part in every stage of the research, from framing the question to acting on the findings. The approach focuses on social and environmental inequities and joins local knowledge with research methods.

Worked example

A school psychologist wants to know whether a points system raises one fourth grader's homework completion. The teacher can record completion every day, there is no comparable group of students to serve as a control, and it would do no harm if completion fell back to its usual level for a short time.

1. Choose a single-case design: with one student and daily measurement, comparing phases within the same child fits better than a group design.
2. Pick the specific design: because completion is expected to drop if the points stop and a brief drop is acceptable, a reversal (ABAB) design suits: baseline, points, withdrawal, points again.
3. Establish a stable baseline: record completion for several days before starting, so any effect of the points shows against a clear starting level.
4. Look for replication: if completion rises with points, falls when they are withdrawn and rises again when they return, the repeated change makes other explanations unlikely; the graph is judged mainly by visual analysis.

A design is chosen to fit the question, the setting and whether the behavior can safely return to baseline.