Scientific working methods: experimental design, data analysis, error treatment, modelling, evaluation
What this domain tests
Scientific working methods covers experimental design, data analysis, error treatment, modelling and evaluation. It is assessed across the whole written paper, often in a part that asks you to describe a measurement, linearise a graph, state uncertainties and judge whether a model is adequate for the data.
Designing an experiment
- State a hypothesis naming the independent variable, the dependent variable and everything to be kept constant.
- Choose instruments by resolution and range; note the reading uncertainty (analogue: half the smallest division; digital: one unit in the last digit).
- Repeat and average to reduce random error; calibrate to reduce systematic error such as zero error or a background count rate.
Analysing data
- Plot graphs with labelled axes, units and a sensible scale; draw a best-fit straight line, not dot-to-dot.
- Linearise to obtain parameters from the slope: T² against l gives slope 4π²/g, so g = 4π²/slope.
- Report the slope with unit and uncertainty, estimated from the steepest and shallowest plausible lines.
Errors and modelling
- Random errors scatter in both directions and are reduced by averaging; systematic errors shift all values the same way.
- Propagate uncertainties: absolute errors add when quantities are added, relative errors add when they are multiplied or divided.
- State the model assumptions (no air resistance, point mass, uniform field) and what each does to the result's direction.
Worked example
A pendulum gives a straight line through the origin with slope 4.05 s²/m. Then g = 4π²/4.05 ≈ 9.75 m/s². Compared with 9.81 m/s² the deviation is about 0.6 %, plausible if the amplitude was a few degrees and l was measured to ±1 mm. A good evaluation names a cause and its direction: the effective length exceeds the measured string by the bob's radius, which pushes g low - exactly the observed sign.
Traps
- Quoting eleven digits from the calculator display.
- Writing "human error" or "measurement error" without naming the measurement and the direction.
- Forcing a line through the origin when the intercept has physical meaning (background, zero error).
- Calling a systematic error random because it differs between groups.