Research: designs, bias, and evidence appraisal
The research domain tests whether you can read, critique, and design studies well enough to ground practice in evidence. Items focus on design–question fit, bias, statistical interpretation, and human-subjects ethics.
Core rules
- Design fits the question: cross-sectional describes prevalence without temporality; case-control suits rare outcomes and yields odds ratios; cohort yields incidence and relative risk; randomised trials support causal inference.
- Hypotheses are null and alternative. Type I error is a false positive (alpha), Type II a false negative (beta); power equals 1 minus beta.
- A p-value is the probability of data at least this extreme if the null hypothesis is true. It is not the probability the null is true, and it says nothing about the size or importance of an effect. Confidence intervals convey precision and magnitude.
- Bias types to name: selection, information/measurement, recall, social desirability, observer, and confounding. Randomisation addresses confounding; blinding addresses observer and measurement bias.
- Internal validity supports causal claims; external validity supports generalisation. Efficacy trials answer 'can it work'; effectiveness studies answer 'does it work in routine conditions'.
- Ethics: IRB review, informed consent that is voluntary, competent, informed, and comprehended; extra protections for vulnerable populations; minimal risk; data security; accurate reporting. Programme evaluation can still be human-subjects research.
- Qualitative rigour: credibility, transferability, dependability, confirmability, achieved through triangulation, member checking, audit trails, and saturation.
- Evidence hierarchy places systematic reviews and meta-analyses above single studies, but study quality can invert the ranking.
Worked example
A 40-person pilot reports p = 0.04 with a 0.3 kg/m² BMI reduction. Correct reading: the result is statistically significant, clinically trivial, and the small sample leaves it underpowered for subgroup conclusions. The defensible conclusion is 'promising, warrants a larger trial', not 'the programme works'.
Traps that cost marks
- Converting correlation into causation.
- Reading p = 0.04 as a large or important effect.
- Generalising from a convenience sample.
- Skipping ethics review because the work is 'just evaluation'.
- Confusing odds ratios with relative risk.