A pupil records two temperatures:
- Start = 70°C
- End = 58°C
Then the worksheet asks for the temperature decrease.
The pupil calculates 12°C—and erases the original two readings because “12°C is the answer we need”.
The final number is correct.
The evidence trail is now weaker.
Raw data tell us what was observed or measured. Derived values tell us what we calculated from those observations. Good Science keeps both.
This guide belongs to the Primary 4 Science Learning Hub. It develops a distinct data-integrity habit: preserve the original record, show the calculation that creates a new value, and make it possible for another learner to trace the conclusion back to the evidence.
The goal is not advanced data science. It is transparent school Science.
Quick Answer: The Evidence-Trail Loop
OBSERVE / MEASURE → RECORD RAW VALUE → LABEL CONDITION + UNIT → CALCULATE DERIVED VALUE → KEEP THE WORKING → INTERPRET → CONCLUDE → RETAIN THE ORIGINAL
This is an eduKate teaching routine, not an official MOE marking formula.
Wait, What? The Calculation Is Not the Observation
Start temperature = 70°C.
Final temperature = 58°C.
Temperature decrease = 12°C.
Only the first two numbers were directly read from the measuring instrument.
The 12°C value was calculated.
All three are useful—but they have different jobs.
1. What is raw data?
For this P4 guide, raw data means the original observation or measurement recorded during the investigation before the learner transforms it into another value.
Examples:
- 70°C at the start;
- 58°C after 15 minutes;
- 14 cm shadow width;
- 100 mL before pouring;
- 100 mL after pouring;
- Plant A height = 12 cm on Day 1;
- wilting score = 2 using a stated classroom scale;
- “leaves drooping” as a written observation.
Raw does not mean “messy”.
It means closest to the original observation event.
2. What is a derived value?
A derived value is produced from one or more raw values through a stated calculation or transformation.
Examples:
- temperature decrease = start − end;
- growth = final height − initial height;
- object volume by displacement = final water reading − initial water reading;
- difference between two shadow widths;
- range of repeated readings = largest − smallest;
- average, if and when the teacher explicitly teaches and requires it.
The calculation should never become indistinguishable from the original measurement.
3. Raw and derived data need different labels
| Evidence type | Example | How obtained |
|---|---|---|
| Raw measurement | 70°C | read thermometer at start |
| Raw measurement | 58°C | read thermometer after 15 min |
| Derived value | 12°C decrease | 70 − 58 |
A table like this prevents the learner from saying “we measured a 12°C decrease directly” when the decrease was actually calculated from two measurements.
4. Why preserve the original readings?
Because another person may need to:
- check the calculation;
- spot a copied value;
- identify a unit error;
- compare final values separately from changes;
- recalculate after a correction;
- judge whether two conditions started fairly;
- explain an anomaly.
When the raw values disappear, these checks become harder or impossible.
5. Evidence trail: from world to conclusion
A strong evidence trail can be shown as:
WORLD / SET-UP → OBSERVATION → RAW RECORD → CALCULATION → DERIVED VALUE → INTERPRETATION → CONCLUSION
Each step should be understandable.
If the learner jumps from “Cup A was foam-wrapped” straight to “Foam is always best”, the evidence trail has missing links.
6. Worked case | Cooling
| Condition | Start / °C | End / °C | Decrease / °C |
|---|---|---|---|
| Foam | 70 | 61 | 9 |
| Cloth | 70 | 57 | 13 |
Raw data:
- 70, 61, 70, 57.
Derived data:
- 9, 13.
Interpretation:
Foam condition showed the smaller temperature decrease.
Conclusion:
Under these tested conditions, foam reduced cooling more than cloth.
Why the raw values matter
If only 9°C and 13°C remain, we lose immediate proof that both cups started at the same temperature.
The original readings support both the calculation and the fairness of the comparison.
7. Worked case | Displacement
Initial water volume = 42 mL.
Final water volume with object submerged = 58 mL.
Derived object volume:
58 − 42 = 16 mL.
If the learner keeps only “16 mL”, another person cannot check whether the subtraction was performed correctly.
8. Worked case | Plant growth
| Day | Height / cm |
|---|---|
| 1 | 12 |
| 7 | 17 |
Derived change:
17 − 12 = 5 cm.
Do not overwrite Day 1 with “growth = 5 cm”.
Starting state and change answer different questions.
9. Worked case | Shadow
At 10 cm object–torch distance:
shadow width = 18 cm.
At 20 cm:
shadow width = 14 cm.
Derived difference:
18 − 14 = 4 cm.
The 4 cm difference is useful, but the original pair is still necessary to see direction: which condition had the larger shadow?
10. Raw observation can be words, not only numbers
Observation:
“Several leaves are drooping.”
Possible derived classroom category:
wilting score = 2 using a stated eduKate scale.
The category is derived from a visible observation according to a rule.
Keep the rule and the original description when the judgement is important.
11. Tables should separate measured and calculated columns
Good table design makes provenance visible.
Example headings:
- Initial temperature / °C
- Final temperature / °C
- Temperature decrease / °C
The third column is not another direct thermometer reading.
Its heading should make the calculation job clear.
12. Do not “clean up” raw data by rewriting history
A reading is 14 cm.
The learner later thinks it “should have been 15”.
Do not silently replace 14 with 15.
If there is a confirmed transcription error—for example the original photographed ruler clearly shows 15 but the table says 14—make a transparent correction.
Keep enough information to show:
- original entry;
- reason for correction;
- corrected value;
- which derived values changed as a result.
13. A correction is different from a new measurement
If the learner mis-copies 58°C as 85°C, correcting the transcription is not the same as re-measuring the water later.
One repairs the record.
The other creates a new observation under a different time condition.
Do not merge them.
14. Keep working visible
Instead of:
Decrease = 12°C.
Write:
70°C − 58°C = 12°C.
The working shows which values were used.
This makes mistakes easier to locate.
15. Calculator output is not new evidence
A calculator can transform numbers.
It does not add observations.
If the raw readings are wrong, a perfect calculation can still produce a misleading answer.
Therefore:
calculation accuracy cannot rescue poor measurement evidence.
16. A graph is processed presentation, not a replacement for the data table
A graph reorganises data visually.
It helps patterns become visible.
But if the graph is drawn incorrectly, the original table should allow the mistake to be found.
Keep the raw or source table even after creating the graph.
Use Constructing Tables, Graphs and Data Displays for graph mechanics.
17. A photograph can be part of the raw record
A photograph can preserve:
- apparatus arrangement;
- scale position;
- visible plant condition;
- time-stamped context.
But any measurement derived later from the image should remain labelled as derived from the photograph.
If the image is resized or cropped, preserve the original.
18. A data logger creates raw readings—and later calculations
Batch 20’s Data Loggers, Sensors and Automatic Measurements explains automated records.
The logged values are raw device readings for that run.
A later difference, range or summary is derived.
Do not delete the original log after making a prettier chart.
19. Raw data and honest precision
Raw values should be recorded at the precision actually supported by the instrument.
Do not first invent extra decimals and then calculate with them.
Batch 22’s Small Differences, Scale Intervals and Honest Precision develops this.
20. Raw data and missing data
A blank is not raw data.
It is an absence in the record whose reason needs explanation.
Do not replace it with zero unless zero was actually measured.
Use Missing Data, Blanks, Zeros and Exclusions for that distinction.
21. Raw data and “no change”
Suppose start and end readings are both 15 cm.
Raw data:
- 15 cm;
- 15 cm.
Derived change:
0 cm at the recorded resolution.
Interpretation:
no detectable change with this method.
Do not jump from zero derived change to “nothing happened”.
22. Raw data and anomalies
Repeated readings:
- 14 cm;
- 15 cm;
- 29 cm;
- 14 cm.
The 29 cm result is raw evidence until a reason is found.
Do not erase it.
Annotate the record:
“Trial 3 unusual; set-up checked and repeated.”
If later excluded from a summary because the object was confirmed to have moved, preserve the original trial and reason.
23. Raw data and sample selection
Raw data can still be biased if only convenient specimens were measured.
Data integrity begins before recording.
The Batch 21 guide on Sampling, Representative Cases and Avoiding Cherry-Picking asks which cases became data in the first place.
24. Raw data and observer expectations
A recorded value may reflect a subjective boundary choice.
Keep the measurement definition with the record.
Otherwise “14 cm” looks objective while hiding how the edge was chosen.
25. The Evidence-Trail Card
| Step | Question | Check |
|---|---|---|
| Raw | What did we directly observe or measure? | □ |
| Context | Which condition, time and unit belong to it? | □ |
| Derived | What calculation created the new value? | □ |
| Working | Can another learner reproduce the calculation? | □ |
| Correction | Was anything changed after recording, and why? | □ |
| Interpretation | What pattern or relationship does the value support? | □ |
| Conclusion | Does the claim stay within the evidence? | □ |
26. Original Evidence-Trail Casebook
Case 1 | Cooling
Raw: 70°C, 61°C.
Derived: 9°C decrease.
Lesson: keep both states and the calculated change.
Case 2 | Plant growth
Raw: Day 1 = 12 cm, Day 7 = 17 cm.
Derived: +5 cm.
Lesson: change is not a replacement for the baseline.
Case 3 | Displacement
Raw: 42 mL, 58 mL.
Derived: 16 mL object volume.
Lesson: calculated property depends on two measured values.
Case 4 | Wrong subtraction
70 − 58 accidentally written as 22.
Repair: visible working exposes the arithmetic error.
Case 5 | Corrected transcription
Notebook photograph shows 58; table says 85.
Repair: correct transparently and recalculate affected derived values.
Case 6 | Graph without table
Plotting error suspected but source data discarded.
Weakness: evidence trail broken.
Case 7 | Calculator decimals
Raw data whole degrees; calculator prints 12.000000.
Repair: report consistent precision.
Case 8 | Missing reading
Blank at 10 minutes changed to zero.
Wrong: absence of data is not a zero measurement.
Case 9 | Excluded trial
Spill invalidates volume comparison.
Good practice: preserve trial, mark spill, exclude from specific summary with reason.
Case 10 | Photo-derived height
Height calculated from a photograph with scale.
Label: derived from image, not directly read from plant at the later analysis time.
27. Corrections should leave a trace
On paper:
cross out the wrong entry once, keep it readable where school practice allows, write the corrected value and add a brief reason if important.
Digitally:
preserve the original file or make a copy before editing.
Do not silently rewrite history.
This habit is simple enough for Primary 4 and foundational for later scientific integrity.
28. Derived values should name the operation
Instead of:
“12°C.”
Write:
“Temperature decrease = 70°C − 58°C = 12°C.”
Instead of:
“5 cm.”
Write:
“Growth in measured height = 17 cm − 12 cm = 5 cm.”
The label prevents a reader from confusing final state, change and difference.
29. Not every processed value deserves to be calculated
A child sees three repeated readings and wants to average them because “scientists use averages”.
If averaging has not been taught or does not help the question, do not add calculation for decoration.
At Primary 4, often the better move is to:
- show all readings;
- describe their consistency;
- identify an unusual value;
- compare ranges or simple differences when appropriate.
30. The evidence trail protects against answer-key thinking
If only the final conclusion matters, the child may learn to chase a sentence.
If the evidence trail matters, the learner must understand:
- where the data came from;
- which transformations were made;
- why the conclusion follows.
This makes Science less about guessing the teacher’s preferred phrase and more about producing inspectable reasoning.
31. Original Practice Set
- What is raw data?
- What is a derived value?
- Why should original readings be preserved?
- Is a temperature decrease directly measured or calculated?
- Why should table headings distinguish raw and derived columns?
- What is wrong with erasing 70°C and 58°C after calculating 12°C?
- Why should visible working be kept?
- What is the difference between correcting a transcription and taking a new measurement?
- Why is a calculator output not new scientific evidence?
- Why should a graph not replace the source table completely?
- What should happen to an anomalous raw reading?
- Can raw data still be biased?
- How should a missing value be recorded?
- Why should an excluded trial remain documented?
- What is the evidence trail from measurement to conclusion?
- Why should derived values inherit honest precision from raw data?
- What does a zero derived change mean?
- Why may a qualitative observation be considered raw data?
- When might a photograph be raw evidence?
- Write one complete evidence-trail sentence for plant growth.
32. Practice Answers
1. The original observation or measurement recorded during the investigation before later transformation.
2. A value calculated or produced from one or more raw values.
3. They allow calculations, fairness, corrections and later interpretations to be checked.
4. Usually calculated from starting and final temperature readings.
5. So readers know which values came directly from measurement and which were computed.
6. The evidence supporting the calculation and starting-condition comparison is lost.
7. It shows which values were used and exposes arithmetic mistakes.
8. A correction repairs the record of an existing observation; a new measurement creates a new observation under a new time/condition.
9. It transforms existing numbers but does not observe the world.
10. The table is needed to check plotting and preserve exact source values.
11. Keep it, annotate it and investigate rather than silently delete it.
12. Yes. Biased sampling or expectation-influenced reading can affect what becomes raw data.
13. As missing, blank or not recorded with a reason where known—not as zero unless zero was measured.
14. So the exclusion is transparent and can be reviewed.
15. Observe/measure → raw record → calculation → derived value → interpretation → conclusion.
16. Calculations cannot create measurement detail not present in the source readings.
17. The recorded start and end values were the same at the measurement resolution; it does not prove nothing happened between them.
18. Directly observed words such as “leaves drooping” are original observations even though they are not numbers.
19. When it directly records a scene, set-up or visible observation at a stated time.
20. Example: “Plant A measured 12 cm on Day 1 and 17 cm on Day 7, so the derived increase in measured height was 5 cm.”
33. The Evidence-Trail Diagnostic
| If the learner… | Likely weak link | Repair |
|---|---|---|
| keeps only final answer | provenance | retain raw readings |
| cannot show calculation | derived-value transparency | write operation |
| changes value silently | record integrity | document correction |
| replaces blank with zero | data semantics | preserve missing status |
| uses polished graph only | source retention | keep original table/log |
34. A 40-Minute Evidence-Trail Lesson
Minutes 1–5: classify raw vs derived values.
Minutes 6–10: build a cooling table with separate columns.
Minutes 11–15: calculate changes with visible working.
Minutes 16–20: correct one transcription transparently.
Minutes 21–25: inspect an anomalous reading.
Minutes 26–30: compare raw table and graph.
Minutes 31–35: trace one conclusion backward to evidence.
Minutes 36–40: transfer to plant, shadow or image data.
35. What Parents and Tutors Can Ask
- “Which number did you actually measure?”
- “Which number did you calculate?”
- “Where is the original reading?”
- “Can I reproduce your calculation?”
- “Did you correct anything after recording?”
- “Why was this trial excluded?”
- “Could this conclusion be checked from your table?”
- “What would be lost if we kept only the final answer?”
36. Continue Batch 22
- Small Differences, Scale Intervals and Honest Precision
- No Change, Null Results and What You Can Conclude
- Missing Data, Blanks, Zeros and Exclusions
The Quiet Return
Science does not become more trustworthy by hiding the steps between observation and answer.
Keep what you saw. Keep what you measured. Show what you calculated. Mark what you corrected. Then let another learner walk the same trail from evidence to conclusion.
