S2.2.3—Data quality
- Syllabus
- First assessment 2025
- Objective
- S2.2.3
- Level
- SL
Assess Data Quality
Data can be precise without being accurate—and accurate-looking data can come from an invalid test
Assess each quality separately against the intended measurement and research question.
| Quality | Question to ask | Evidence or improvement |
|---|---|---|
| Accuracy | How close is the result to an accepted or well-supported value? | Calibration, correction of systematic offset or comparison with a reference |
| Precision | How closely do repeated readings agree, or how small is the measurement resolution/uncertainty? | Smaller spread and finer justified resolution |
| Reliability | Are results consistent across sufficient repeats or repeated trials? | Repeat under the same conditions and compare the pattern |
| Validity | Does the method isolate and measure the intended relationship? | Appropriate controls, range, method and interpretation |
Repeats target random variation
Repeating a drop-time measurement under the same conditions can reveal spread and improve precision of a representative result. Saying only “find an average” is incomplete: the reason is to reduce the influence of random variation, not to guarantee accuracy.
Systematic effects need a changed method
A zero offset can shift every reading together, so tightly clustered repeats may still be inaccurate. Calibration or correction can address the offset; averaging the same biased method cannot. Validity can also fail even when readings are precise and reliable if another variable causes the observed change.
S2.2.3 Exam Analysis
Assessment in practice
1 marksHow it is assessed
Questions explain why one value per condition is poor or identify systematic error from a graph.
Command terms
Suggest / Identify
What earns marks
Mention random variation/outliers and the need for multiple measurements, or identify the non-origin trend as systematic evidence.
Watch for
Saying one trial is poor only because it is less precise, without linking it to random variation or outliers.
Retrieve Data Evaluation
Record before judging
Keep qualitative observations, labelled quantitative data, units and uncertainty visible.
Test the model
Use the graph shape, intercept, fit, error bars and repeats to decide whether a relationship is supported and whether errors are random or systematic.