Research Design, Biostatistics, and Evidence Evaluation
Key Concepts
- Study designs: Observational (cross-sectional, cohort, case-control), experimental (randomized controlled trials), and quasi-experimental.
- Bias: Selection bias, information bias, confounding, recall bias.
- Validity: Internal (causality) vs. external (generalizability).
- Statistics: Descriptive (mean, median, SD), inferential (t-tests, ANOVA, chi-square, regression), effect size, p-values, confidence intervals.
- Evidence hierarchy: Systematic reviews, RCTs, cohort studies, case-control, case reports, expert opinion.
- Critical appraisal: Assessing study quality, relevance, and applicability.
Traps
- Confusing statistical significance with clinical relevance.
- Assuming correlation implies causation.
- Overlooking confounding variables in observational studies.
- Misinterpreting p-values (not the probability that the null is true).
- Failing to consider sample size and power.
Worked Example
A study reports that dogs fed a grain-free diet have more aggression (p=0.04). Before accepting, check: Was it a randomized trial? What confounders (e.g., owner education, breed) were controlled? Rule of thumb: a single study is not definitive; consider the body of evidence and study design.