From question to defensible evidence
Physics builds models, measures the world and states honestly what the data can support. A result without uncertainty is incomplete.
Learning goals
- design a fair test with independent, dependent and controlled variables;
- state model assumptions and limits;
- distinguish random scatter from systematic bias;
- report and propagate measurement uncertainty;
- test a model with linearisation and residuals.
Investigation cycle
1. Ask and predict
Turn a broad question into a testable relationship. A hypothesis identifies what changes, what is measured and why the model predicts the pattern.
2. Measure fairly
Change the independent variable, observe the dependent variable and keep important controls constant. Record units, instrument resolution and conditions; repeat every reading.
3. Analyse and revise
Graph the data, fit a justified model, inspect residuals and compare disagreement with uncertainty. State both the supported conclusion and its limits.
Model, validity and reliability
A model keeps the features needed for a question and deliberately omits others. The simple-pendulum model $T=2\pi\sqrt{L/g}$ assumes a small angle, a point-like bob, a light inextensible string and negligible drag. It is useful inside those limits. Validity asks whether the method tests the intended relationship; reliability asks whether repeats agree. A biased procedure can be reliable but inaccurate.