Guide
How to read a new study without overclaiming
A compact method for separating a result, its design, its uncertainty and the claim a reader can safely repeat.
Find the actual study
Start with the paper, preprint or dataset, not only the press release. Record whether it has been peer reviewed and what population or system was studied.
Ask what was compared
Identify the baseline, sample size, outcome measure and time window. A percentage without its denominator or comparison is not enough to interpret the result.
Keep the limit attached
Separate what the experiment shows from what it might imply elsewhere. Look for uncertainty intervals, missing controls, replication and conflicts of interest.
What this guide does not cover
This does not replace specialist review of a clinical, financial or safety decision.
Links checked
- arXiv — https://arxiv.org/checked
- PubMed — https://pubmed.ncbi.nlm.nih.gov/checked