SO5005 Chap.3 Critical Appraisal and Research Ethics
Critical Appraisal and Research Ethics
Validity comes before precision
Critical appraisal tests whether a result deserves belief, how large and precise the effect is, and whether it transfers to a new setting. Selection bias changes who enters, information bias changes measurement, and confounding supplies a rival explanation.
Random error creates instability rather than a systematic direction.
Ethics strengthens design
Respect, beneficence and justice shape recruitment, consent, burden and distribution of benefit. Privacy concerns access to the person; confidentiality concerns entrusted information.
Unnecessary identifiers, coercive recruitment and unclear withdrawal are ethical weaknesses that often reveal an unfocused scientific plan.
What this chapter covers
- 01
Internal validity
- 02
Selection and information bias
- 03
Confounding and random error
- 04
Effect size and precision
- 05
Consent and voluntariness
- 06
Privacy and confidentiality
Worked example · free
A precise estimate from a narrow sample
- 2Locate the sampling route.
- 2Separate precision from selection.
- 2Propose a repair.
Key terms
- Selection Bias
- Systematic distortion created by who enters or remains in a study.
- Information Bias
- Systematic distortion created by inaccurate or unequal measurement.
- Confidentiality
- Protection governing how information entrusted to a researcher is stored, used and disclosed.
Critical Appraisal and Research Ethics FAQ
Does a narrow confidence interval prove validity?
No. A narrow interval indicates statistical precision under the analysis assumptions, but a biased sample or measurement process can produce a precise estimate of the wrong quantity. Appraise recruitment and measurement before interpreting interval width.
How should a limitation be written?
Name the design feature that creates the limitation, explain which estimate or conclusion it could change, and state the likely direction only when justified. Then offer a feasible repair or sensitivity check instead of attaching a generic limitation list.
Are ethics and methods separate sections only?
They may appear under different headings, but the decisions overlap. Recruitment affects justice and selection, unnecessary data affect confidentiality and focus, and burdens without scientific value fail beneficence. Ethical reasoning should therefore be visible throughout the method.
Assessment move
Appraise one paper with three passes: validity, magnitude and transfer. For every concern, write the causal path from the design choice to the distorted result. Finish by identifying one ethical safeguard that also improves evidence quality. During the validity pass, reconstruct recruitment, exposure measurement, outcome measurement and follow-up as a flow rather than reading the discussion section first.
Mark where group membership could influence entry, measurement or loss. For confounding, draw a third factor pointing toward both the exposure and outcome; if you cannot state both arrows, the label is premature. During the magnitude pass, copy the estimate, interval, unit and denominator, then distinguish a wide interval from a biased centre.
Ask whether the effect size would matter at the smallest and largest values still compatible with the interval. During the transfer pass, compare the study population, care pathway, timing and measurement with the setting in which a decision will be made. Do not reject transfer because two locations have different names; identify the mechanism that could make the effect differ.
Build an ethics table with columns for participant burden, expected value, voluntary choice, privacy, confidentiality and fairness of recruitment. Every protection should be operational: who gives information, when consent can be withdrawn, which identifiers are collected, who can access them and when they are removed. End with a repair, not a criticism list.
State the smallest change that would reduce the main threat, and explain which part of the conclusion becomes stronger. This makes appraisal useful for designing your own proposal rather than merely finding faults in someone else's work.
Working through Critical Appraisal and Research Ethics in SO5005? Sia is AskSia’s AI Statistics tutor — ask any SO5005 Critical Appraisal and Research Ethics question and get a clear, step-by-step explanation grounded in how SO5005 is taught and assessed. Read this chapter free, then take your hardest questions to Sia.
