Analysis, communication and evaluation
- Syllabus
- 2024
- Topic
- —
- Level
- —
Learning objectives
—Analyse data and draw conclusionsUse scientific knowledge and understanding to analyse and interpret data and draw conclusions consistent with experimental evidence.—Communicate experimental findingsCommunicate experimental findings using appropriate technical language, relevant calculations and graphs.—Assess experimental reliabilityAssess the reliability of an experimental activity.—Evaluate accuracy and validityEvaluate data and methods, taking into account factors that affect accuracy and validity.
Turn data into a conclusion the evidence can support
A sound conclusion is the last link in an evidence chain: identify the pattern, test it with data, compare the results, then state whether the evidence supports the proposed relationship. The conclusion must match both the direction and the strength of the evidence.
Use this sequence: inspect the full table or graph for trend and anomalies → state the proposed relationship precisely → select at least two well-separated data points → perform the relevant calculation for each point → compare the calculated values → give a qualified conclusion.
Example: test whether distance is proportional to time squared. At 5.0 s, distance/time² = 15/25 = 0.60 m/s². At 10.0 s, distance/time² = 60/100 = 0.60 m/s². The equal values support the proposed relationship over the measured range. Show the substitution, value and unit so the comparison is traceable.
A curve that falls as x rises shows a negative relationship, but does not by itself prove inverse proportionality. One point cannot test constancy, and an anomalous point should be identified and checked rather than silently deleted. Say “supports” rather than “proves” because experimental evidence is limited to the data collected.
Communicate findings so the evidence is easy to follow
Scientific communication lets another reader see what was measured, how it was processed and how it supports the conclusion. Use precise technical language, display calculations in a checkable order and choose a graph treatment that represents the data rather than forcing a preferred pattern.
| Evidence feature | Clear communication |
|---|---|
| axes | label each quantity and unit; use a sensible scale |
| plotted data | mark points accurately and identify any anomaly |
| best fit | use a straight line for a linear trend or a smooth curve for a curved trend, with points reasonably balanced around it |
| calculation | write the relationship, substitute values, give the result and unit |
| conclusion | name the trend and cite the comparison that supports it |
For continuous measurements, use a line graph. Decide between a straight line and smooth curve from the overall distribution of points, not by joining every point dot-to-dot. A justified anomalous point need not pull the best-fit line or curve away from the main pattern.
A graph without quantity-and-unit labels is ambiguous, and a calculation with only a final number cannot be checked. Technical language should increase precision, not hide the reasoning: every reported finding should still connect visibly to a plotted pattern, table value or calculation.
Assess reliability by checking consistency
Reliability is about consistency: would repeated measurements made under the same conditions give similar results? It is judged from repeat readings and their spread, not from whether a single value looks plausible or agrees with an expected answer.
For each condition, take repeated readings → compare their spread → repeat any suspicious result → identify an anomaly only when the repeats provide evidence → calculate a mean from the valid readings. State both the action and its purpose: repeats reveal variation, while the mean reduces the influence of random variation.
Suppose three readings at one setting are 4.8, 4.9 and 7.2. Do not discard 7.2 immediately. Repeat the measurement at that setting and check the method. If the new readings cluster near 4.8–4.9, identify 7.2 as anomalous, exclude it with that justification, and report the mean of the consistent readings.
Repeating once is not enough to establish a pattern of consistency, and averaging an unexamined anomaly can make the result less representative. Repeats and a mean improve reliability, but they do not remove a systematic offset that shifts every reading in the same direction.
Evaluate accuracy and validity as different questions
Accuracy asks how close a measurement is likely to be to the accepted or true value. Validity asks whether the method and evidence genuinely test the intended relationship. A useful evaluation names a specific limitation, explains its effect, proposes a targeted improvement and states why that improvement helps.
| Question | Possible limitation | Targeted improvement |
|---|---|---|
| accuracy | zero error, parallax or scale divisions too coarse | zero or calibrate the instrument, read square-on, or use finer resolution |
| validity | another relevant variable changes | control that variable so only the independent variable changes |
| validity of a claimed graph shape | too few readings or large gaps hide the shape | take more readings at smaller intervals, especially between existing points |
If the evidence cannot distinguish a straight line from a curve, “take more readings” is incomplete. Specify smaller intervals in the uncertain region: the extra points reveal whether the gradient stays constant or changes, so the conclusion about the relationship is better supported.
More repeats mainly test reliability; they do not automatically correct a zero error or make an uncontrolled comparison valid. Likewise, a wider range or smaller intervals can strengthen the validity of a relationship claim without making each individual instrument reading more accurate. Match every improvement to the limitation it addresses.