Research reports findings with their methods and limits attached: what a study shows, in whom it was measured, and what remains open.
Sampling error is the only uncertainty a poll's +/- number covers; weighting, mode and nonresponse often move results further than the formula suggests.
Observational studies and randomized trials answer different questions, and the most famous reversal in nutrition research shows exactly why.
A study too small to detect the effect it hunts will usually miss it, and when it does not, the effect it reports is inflated.
When 270 psychologists tried to repeat 100 famous studies, most effects shrank or vanished — and psychology rebuilt its methods from the wreckage.
Most biomedical and physics research now appears online before peer review, so readers need a working rule for how much weight a preprint can carry.
Forest plots and pooled estimates look authoritative, but the strength of a meta-analysis depends on choices made long before the statistics run.
The most quoted number in science answers one narrow question, and misreading it has shaped a decade of debate about how research should be judged.