Research
What you must know
Research questions demand design literacy. Know the sequence: review the literature, state a clear question and hypothesis, define variables operationally so they can be measured, select a design, sample, collect and analyze data, and interpret results within limits.
Know designs: descriptive and correlational, cross-sectional, case-control, cohort, quasi-experimental, randomized controlled trial, and qualitative traditions (ethnography, phenomenology, grounded theory, case study). Know the evidence hierarchy: systematic reviews and meta-analyses at the top, then randomized trials, cohort, case-control, cross-sectional, case reports, and expert opinion - while remembering that the right design for a specific question can outrank a misapplied one.
Know sampling: probability (simple random, systematic, stratified, cluster) versus nonprobability (convenience, purposive, snowball, quota). Nonprobability sampling limits generalizability.
Know bias: selection, measurement or instrument, recall, social desirability, interviewer, Hawthorne, and volunteer bias - and how randomization, blinding, standardized instruments, and triangulation counter them.
Traps: reading a p value as the probability the hypothesis is true or as an effect size (p addresses chance only); equating statistical with practical significance; treating correlation as causation; collecting data before IRB approval or informed consent; generalizing from a convenience sample.
Rule of thumb: A confidence interval gives the plausible range of the effect; a p value only says the result is unlikely under the null. Prefer the interval when judging practical value.
Worked example: A cessation study reports a 2-percentage-point improvement in quit rates, p = .04, 95% CI [0.3, 3.7]. It is statistically significant, but the decision-maker must still ask whether 2 points justifies the cost per participant.