Testing and uncertainty
Scientific claims are evaluated by comparing predictions with data while separating measurement error, model uncertainty and theoretical approximation.
PredictionA useful theory should say what should happen under specified conditions.
UncertaintyNo measurement is infinitely precise; uncertainty must be stated.
Falsification in practiceDisagreement may challenge a theory, but one must first exclude experimental and modeling errors.
ConvergenceConfidence increases when independent methods lead to compatible results.