How to perform in the new G2 SEC Science examination with data requires more than reading numbers correctly. Raw readings often need to be transformed before they answer the scientific question. A final value may need to become a change. Two readings may need to become a rate. Repeated measurements may need a mean. Several categories may need a ratio or percentage. A table may need to become a graph. Each transformation changes the representation, not the underlying meaning.
This forty-eighth Learner’s Guide focuses on Science data transformations: raw reading → change → rate → mean → ratio → graph → conclusion. The central rule is: transform only when the new quantity answers the task, and keep units and meaning attached at every step.
For 2027, SEAB lists G2 Science as K223 Science (Physics, Chemistry), K224 Science (Physics, Biology) and K225 Science (Chemistry, Biology). Use the SEAB 2027 G2 syllabus directory for current official details. The numerical examples below are original teaching data rather than official examination questions or real experimental results.
Raw data is not always the answer
A thermometer reading of 54°C is a raw measurement. If the question asks how much the sample cooled from 70°C, the required quantity is the change: 16°C.
The transformation depends on the scientific question. Do not calculate a change merely because two numbers are available.
Transformation one: difference
Difference compares two values by subtraction.
It is useful for temperature change, mass change, distance difference, score difference, concentration difference and many other quantities.
Direction matters
A change can be written as final − initial, which preserves sign, or as a decrease/increase stated as a positive magnitude, depending on the task.
Do not mix the conventions. If temperature falls from 70°C to 54°C, final − initial = -16°C, while the temperature decrease is 16°C.
Name the transformed quantity
Write “temperature decrease = 16°C”, not just “16”.
The label prevents a transformed value from being mistaken for a raw reading.
Transformation two: rate
A rate relates change to time or another quantity.
Examples include speed, reaction progress per second, cooling per minute, flow per unit time or biological change per day.
Average rate
Average rate over an interval uses total change divided by total interval.
It does not necessarily describe the instantaneous rate at every moment inside the interval.
Rate example
Gas volume rises from 10 cm³ to 34 cm³ over 20 seconds. The average increase is 24 cm³ over 20 s, or 1.2 cm³/s.
Do not report 34 ÷ 20 unless the question asks for average volume from zero time and the initial volume is actually zero.
Rate direction
If the measured quantity decreases, the signed rate may be negative. Some questions instead ask for the rate of decrease as a positive magnitude.
Follow the wording and label the result.
Transformation three: mean
A mean summarises repeated numerical values by adding them and dividing by the number of values.
It is useful when the measurements are comparable and averaging is scientifically meaningful.
Mean is not automatically better
A mean can hide variation or an anomaly.
Before averaging, inspect the readings. If one value results from a known procedural mistake, the question may require evaluation rather than blind inclusion.
Mean example
Repeated times are 19.8 s, 20.1 s and 20.0 s. Mean = 59.9 ÷ 3 ≈ 19.97 s before appropriate final reporting.
The mean summarises the readings; it does not make the instrument more accurate automatically.
Mean and systematic bias
If every reading contains the same known offset, the mean contains that offset too.
Averaging addresses variation, not every kind of systematic error.
Transformation four: ratio
A ratio compares quantities multiplicatively.
It can reveal relative proportions when an absolute difference is not the most meaningful comparison.
Ratio example
If two samples have masses 30 g and 20 g, their mass ratio is 3:2.
This is different from saying one is 10 g heavier. Difference and ratio answer different questions.
Transformation five: percentage
A percentage expresses a quantity relative to a base of 100.
The base must be identified. Percentage change, percentage of total and percentage difference are not interchangeable phrases.
Percentage of total
If 12 of 30 samples show a feature, the percentage is 12/30 × 100% = 40%.
The denominator is the total relevant sample, not an arbitrary nearby quantity.
Percentage change
If a measurement increases from 20 to 25, the increase is 5 and the percentage increase relative to the original is 5/20 × 100% = 25%.
The original value defines the base.
Transformation six: normalisation
Sometimes quantities are divided by size, mass, area or another reference so fairer comparison is possible.
Examples include per gram, per unit area or per participant. The normalisation changes the question from total amount to amount relative to a reference.
Normalisation example
Sample A produces 40 units from 20 g; Sample B produces 45 units from 30 g. B has the larger total, but A has 2 units/g compared with B’s 1.5 units/g.
Which comparison matters depends on the scientific aim.
Transformation seven: graphing
A graph transforms a table into a visual relationship.
The values do not change, but patterns become easier to see: increase, decrease, plateau, optimum, anomaly or approximate proportionality.
Choose axes from variables
Independent variable usually goes on the horizontal axis and dependent variable on the vertical axis in ordinary experimental graphs.
The learner should still follow the exact question and context rather than memorising a rule without understanding.
Choose scale to reveal the data
A graph scale should use the plotting area effectively without distorting relationships.
Unequal or unusual scale intervals must be read carefully.
Graph transformation does not create new data
A smooth line of best fit summarises a trend. It does not turn approximate measurements into exact values.
Values read between measured points are estimates supported by the graph representation.
Transformation eight: gradient
Gradient expresses change in y divided by change in x.
Its scientific meaning depends on the variables and units on the axes.
Gradient units
If y is distance in metres and x is time in seconds, gradient has units m/s.
Units help the learner interpret what the gradient represents.
Transformation nine: reciprocal
In some relationships, taking a reciprocal can produce a useful alternative quantity, such as period and frequency where relevant to the syllabus context.
Do not transform merely to make numbers look simpler. The reciprocal must have scientific meaning.
Transformation ten: combining repeats and rate
A learner may need to calculate a rate for each trial, then average the rates, or average measurements before calculating a rate.
These procedures can give different results. Follow the scientific purpose and question instructions.
Order of transformations matters
Transformations are operations, and operations do not always commute.
Averaging percentages, taking percentages of averages and averaging raw values before transformation can produce different answers.
Keep raw data available
Do not replace the original readings with transformed values during practice review.
Raw data allows the learner to check the transformation and inspect anomalies.
The raw-to-derived chain
- raw reading;
- calculated change;
- rate or ratio;
- summary statistic;
- graphical representation;
- scientific conclusion.
Not every task uses every stage. Use only the transformations needed for the question.
Physics data transformations
Physics often uses rates, gradients, ratios and unit conversions.
The learner should keep physical quantities and units visible so the transformed value remains interpretable.
Chemistry data transformations
Chemistry may use temperature change, gas volume change, rate, concentration-related quantities, means or graph trends depending on the syllabus topic.
Separate observation from calculation: a measured colour change is not automatically a numerical concentration unless a valid relationship is provided.
Biology data transformations
Biology may use means, percentages, rates, ratios and normalised comparisons.
Natural variation makes it especially important to inspect spread and sample context before treating one mean as the whole story.
The same raw data can answer different questions
A table of mass over time can support final mass, mass change, average rate of change, percentage change and a graph.
The command word and scientific aim determine which transformation is appropriate.
Do not calculate everything
Over-calculation wastes time and can create irrelevant answers.
Ask what transformed quantity is needed before reaching for a formula.
The transformation cue
Use: What quantity is the question really asking for?
Name it before calculating.
Worked data case one: temperature change
Three samples begin at 65°C and finish at 58°C, 54°C and 49°C after the same interval. The raw final temperatures are not yet the temperature decreases.
The decreases are 7°C, 11°C and 16°C. If the question asks which sample cooled most, compare the decreases. If it asks which ended hottest, compare the final temperatures. Same table, different transformation.
Worked data case two: rate from two readings
Gas volume increases from 8 cm³ at 10 s to 32 cm³ at 30 s.
Change = 24 cm³ over 20 s, so average rate over that interval is 1.2 cm³/s.
Using 32 ÷ 30 answers a different quantity because the interval did not begin at zero volume or zero time.
Worked data case three: mean and anomaly
Repeated lengths are 12.1 cm, 12.0 cm, 12.2 cm and 15.8 cm.
Before averaging, inspect the unusual 15.8 cm value. The question may require a repeat or evaluation. Blindly averaging can hide a possible anomaly.
Worked data case four: percentage of total
In a sample of 50 organisms, 18 show a feature. Percentage = 18/50 × 100% = 36%.
If the question instead asks the ratio showing feature : not showing feature, the answer is 18:32, simplified to 9:16. Percentage and ratio encode related but different representations.
Worked data case five: normalised output
Two samples produce outputs of 72 units and 90 units. Their masses are 24 g and 45 g.
Raw output favours the second sample. Output per gram is 3 units/g for the first and 2 units/g for the second. Which comparison is relevant depends on the aim.
Worked data case six: graph gradient
A graph shows distance increasing from 20 m at 4 s to 50 m at 10 s.
Gradient = (50 − 20)/(10 − 4) = 30/6 = 5 m/s. The numerator and denominator come from changes, not raw endpoint values.
Worked data case seven: average of rates
Three trials have rates 1.1, 1.2 and 1.0 cm³/s. Mean rate = 1.1 cm³/s.
If the trials cover different time intervals or start/end values, calculating one combined rate from pooled data may produce a different answer. The procedure must match the question.
Worked data case eight: percentage change
A measurement rises from 40 to 52. Change = 12. Percentage increase = 12/40 × 100% = 30%.
The final value 52 is not the denominator. The original value is the base.
Worked data case nine: ratio versus difference
Sample A has concentration 8 units; Sample B has 4 units.
A is 4 units higher, but it is also twice B. Difference = 4; ratio A:B = 2:1. Choose based on the wording.
Worked data case ten: graph interpolation
A graph has measured values at 20°C and 30°C. Estimating a value at 25°C uses interpolation within the measured range.
The estimate inherits uncertainty from both measurements and the assumed trend between them.
Data transformation and evidence strength
Use Vol 0036. A transformed number can make a pattern clearer but does not automatically make the evidence stronger.
A percentage calculated from a tiny sample still comes from a tiny sample.
Data transformation and model fit
Use Vol 0044. Transformations can help compare measurements with model predictions, but the model assumptions and measurement quality still matter.
Transformation and uncertainty
Every calculation built from measurements inherits limitations from those measurements.
A precise-looking rate does not become more reliable than the readings used to calculate it.
Do not over-report precision
If raw data is recorded to limited resolution, transformed values should not be reported with meaningless extra decimal places.
Follow the question and appropriate syllabus conventions for numerical reporting.
Unit conversion before transformation
Sometimes units must be converted before rates, ratios or means are meaningful.
Do not average metres and centimetres together without conversion. Do not divide kilometres by minutes and label the result km/h unless time is converted appropriately.
Unit conversion after transformation
In some cases, it is efficient to calculate first and then convert the final unit.
Choose the route that preserves dimensional consistency and reduces error.
The unit audit
- What unit does each raw value have?
- What operation am I performing?
- What unit should the transformed quantity have?
- Does the final unit make scientific sense?
Units are a check on the transformation itself.
The denominator audit
Rates, ratios and percentages depend heavily on the denominator.
Ask: divided by what, and why? The denominator defines the reference.
The interval audit
For rates and changes, make sure both values refer to the intended interval.
Using a starting value from one interval and an ending value from another can produce a clean but meaningless calculation.
The baseline audit
Percentage changes and normalised comparisons require a clear baseline.
If the baseline is zero, ordinary percentage change may be undefined or inappropriate. Do not force the formula.
The sample audit
Means and percentages should use the correct sample or group.
If one category is excluded, the denominator may change. Keep inclusion rules visible.
The transformation decision tree
- What quantity is asked?
- Is the raw reading already that quantity?
- If not, what relationship creates it?
- What is the correct denominator or interval?
- What units result?
- What does the transformed number mean scientifically?
This prevents formula-first data handling.
Data tables: label derived columns
If you calculate change or rate from raw columns, give the derived column a clear heading and unit.
This makes it possible to distinguish measured and calculated values.
Measured versus calculated data
Measured data comes directly from the instrument or observation. Calculated data is derived from measured values.
Both can be used as evidence, but they should not be confused.
Graph measured or calculated values?
Either can be appropriate depending on the investigation.
A graph of raw temperature against time answers a different question from a graph of cooling rate against time.
The same experiment can yield several graphs
One data set may support graphing total amount, change, rate or normalised output.
The scientific aim decides which graph is useful.
Graph choice and transformation
A graph should make the relevant relationship easier to inspect.
Do not transform data merely because a different graph looks more dramatic.
Mean versus median
Where the syllabus and task require it, different summary statistics can respond differently to extreme values.
The learner should use the statistic asked for or scientifically appropriate to the problem, not default automatically to mean.
Spread matters
Two groups can have the same mean and different spread.
A mean is one transformation, not a complete description of the data.
Relative versus absolute change
Absolute change reports the difference in original units. Relative change compares that difference with a reference value.
A 5-unit change can be large for a baseline of 10 and small for a baseline of 100.
Rate versus total change
A fast process over a short time can have a smaller total change than a slower process acting much longer.
Do not treat rate and total amount as interchangeable.
Final value versus change
A sample can end with a higher final value yet have changed less if it started higher.
Always identify whether the question is about endpoint or change.
The data-transformation error ledger
- wrong baseline;
- wrong denominator;
- wrong interval;
- units not converted;
- mean calculated despite obvious anomaly without review;
- rate confused with total change;
- final value confused with change;
- raw and calculated values mixed in one column;
- graph axes chosen for the wrong quantities;
- precision exaggerated.
These categories direct the next repair.
The raw-data-only drill
Give a table and ask the learner to label which questions can be answered directly and which require transformation.
This builds selection before calculation.
The denominator drill
Provide percentage, rate and ratio questions using the same numbers but different reference quantities.
The learner must state the denominator before calculating.
The transform-and-explain drill
After every calculation, require one sentence: “This value means…”
If the learner cannot explain the number, the transformation may have lost its scientific meaning.
The graph-from-table drill
Create a graph from raw values, then create a second graph from derived changes or rates.
Compare what each graph reveals and what each hides.
The reverse-transformation drill
Given a rate and interval, recover the change. Given a percentage and total, recover the count where appropriate.
Reverse work strengthens understanding of the relationship between raw and derived quantities.
The 20-minute data session
- Five minutes: identify required transformed quantities.
- Five minutes: calculate changes and rates.
- Five minutes: calculate means, ratios or percentages.
- Five minutes: interpret one graph and state a bounded conclusion.
This can be adapted to the learner’s actual G2 Science combination.
A four-week data-transformation build
Week 1 — change and rate
Focus on intervals, direction and units.
Week 2 — mean, ratio and percentage
Focus on denominator and sample definition.
Week 3 — graph and gradient
Move between tables, plots and interpretation.
Week 4 — mixed structured questions
Choose transformations from the task rather than from topic labels.
Use structured-question decomposition
Use Vol 0032. Data transformation is one possible scientific operation between evidence and answer.
Use controlled comparisons
Use Vol 0040. A transformed value is useful only when the underlying groups or conditions are comparable.
Use the Science index
If the learner cannot interpret the transformed quantity because the concept is missing, return to the Complete Science Index.
The PSLE bridge
The earlier rule Evidence Before Explanation remains central. G2 adds a further question: what transformation, if any, makes the evidence answer the task?
Use Examination Craft
For pacing and final checking, continue through the Examination Craft hub. Data calculations should remain targeted rather than consume time through unnecessary transformations.
Final rule
Do not transform data because a formula is available. Transform it because the scientific question requires a new quantity.
Name the target, choose the operation, preserve units, keep the denominator or interval visible and explain what the result means. A good transformation makes the evidence easier to interpret without changing what was actually measured.
Transformation choice clinic one: endpoint or change
Two samples finish at 40°C and 45°C. Without starting values, you can compare final temperatures but cannot compare temperature decreases.
The correct transformation depends on having the required baseline.
Transformation choice clinic two: total or rate
Process A produces 60 units in 30 seconds; Process B produces 70 units in 60 seconds.
B has the larger total, but A has the higher average rate: 2 units/s versus about 1.17 units/s. “Better” cannot be decided until the performance criterion is defined.
Transformation choice clinic three: count or percentage
Group A has 20 successes out of 40; Group B has 24 successes out of 60.
B has more successes, but A has a higher success percentage: 50% versus 40%. Absolute count and relative frequency answer different questions.
Transformation choice clinic four: mean or full pattern
Two groups both have mean 10. One is 9, 10, 11; the other is 2, 10, 18.
The equal means do not imply equal spread. If the question concerns consistency, mean alone is insufficient.
Transformation choice clinic five: ratio or percentage
A:B = 2:3. A as a percentage of the total is 2/5 × 100% = 40%.
The ratio compares parts; the percentage compares one part with the whole.
Transformation choice clinic six: change or percentage change
A quantity rises from 10 to 15; another rises from 100 to 110.
Absolute changes are 5 and 10, but percentage changes are 50% and 10%. The larger absolute increase is not the larger relative increase.
Transformation choice clinic seven: raw output or normalised output
Sample A produces 100 units from 50 g; Sample B produces 90 units from 30 g.
A has higher total output; B has higher output per gram. The scientific purpose decides which matters.
Transformation choice clinic eight: table or graph
A table gives exact values. A graph makes the pattern visible.
Use the table for precise retrieval and the graph for relationships, interpolation and trend. One representation does not replace the other.
Transformation choice clinic nine: individual rate or pooled rate
Two trials can have different durations and total changes. Averaging their rates equally gives each trial equal weight; pooling total change over total time weights by duration.
These procedures answer different summaries. Follow the question and scientific aim.
Transformation choice clinic ten: measurement or proxy
A colour score may be converted into concentration only when a valid relationship or calibration is supplied.
Do not invent a numerical transformation from a qualitative observation without evidence.
Transformation pipelines
Some structured questions require several transformations in sequence.
Example: raw masses → mass change → percentage change → graph → comparison. At each stage, label the new quantity so the chain remains auditable.
Pipeline error one: wrong baseline early
If the percentage baseline is wrong, every later graph and conclusion can be wrong even when calculations are internally consistent.
Check the earliest transformation before inspecting the final graph.
Pipeline error two: unit inconsistency
A rate calculated from grams and minutes cannot later be labelled kg/s without conversion.
Unit errors can propagate silently through derived columns.
Pipeline error three: premature rounding
Rounding a rate early can distort a later mean or percentage.
Keep sufficient precision through the pipeline and round at the final stage when appropriate.
Pipeline error four: derived value treated as raw
A calculated rate may be copied into a table beside measured values without clear labelling.
Always distinguish what the instrument measured from what the learner calculated.
Pipeline error five: graph of the wrong quantity
A graph of final temperatures cannot answer a question about cooling rates unless the necessary time and change information is represented.
Choose axes from the quantity the scientific question actually concerns.
The transformation audit
- Raw values identified?
- Target quantity named?
- Baseline or denominator correct?
- Interval correct?
- Units consistent?
- Derived quantity labelled?
- Precision sensible?
- Conclusion based on the transformed quantity, not a nearby one?
This audit is especially useful in long data-response questions.
Transformations and anomalies
An anomalous raw reading can produce an anomalous rate or percentage.
Do not treat the derived anomaly as a new independent problem. Trace it back to the raw data.
Transformations and repeats
Repeated raw measurements can be summarised, but review whether every repeat represents the same condition.
Do not average across conditions that were meant to be compared separately.
Transformations and controlled comparisons
A beautiful percentage or graph cannot rescue a confounded experiment.
Data transformation improves representation, not experimental design.
Transformations and evidence boundaries
A calculated rate over one interval should not automatically be treated as the rate at every moment.
A percentage from one sample should not automatically describe a larger population.
Transformations and model assumptions
A linear graph may justify using a gradient over the measured range, but extrapolation beyond that range assumes the relationship continues.
Use the model-data discipline from Vol 0044 when extending transformed trends.
The no-unnecessary-transformation rule
If the question asks for final temperature, give final temperature. Do not calculate percentage cooling unless needed.
Every extra operation creates another opportunity for error.
The one-transformation-at-a-time rule
In difficult questions, write each transformed quantity on its own line before using it again.
This makes the chain easier to check and repair.
The explain-the-number rule
After deriving a rate, mean, percentage or ratio, write what it represents.
A number without interpretation can be mathematically correct but scientifically incomplete.
The graph-label rule
Axis labels should identify quantity and unit. A graph titled only “Results” gives too little information.
Clear labels preserve meaning through the visual transformation.
The table-heading rule
Derived columns need descriptive headings such as “temperature decrease / °C” or “average rate / cm³ s⁻¹” where appropriate.
This prevents a column of calculated values from being mistaken for raw readings.
The denominator-first habit
Before percentage, ratio or rate, state the denominator in words.
“Per second”, “out of total”, “relative to original”, “per gram”. The words clarify the reference before arithmetic begins.
The baseline-first habit
For percentage change, state the original or reference value explicitly.
This avoids the common error of dividing by the final value.
The interval-first habit
For rate, state start and end times or positions before subtracting.
This prevents mismatched endpoints.
The mean-first question
Before calculating a mean, ask whether the values are comparable repeats of the same quantity under the same condition.
If not, averaging may destroy the comparison the experiment was designed to make.
The Science data error ledger
- wrong transformed quantity selected;
- wrong baseline;
- wrong interval;
- wrong denominator;
- units inconsistent;
- mean hides an anomaly without review;
- raw and derived values confused;
- graph uses the wrong variable;
- derived value overgeneralised;
- precision exceeds the measurements.
These categories make data mistakes easier to diagnose.
The data-choice drill
Give one table and five possible questions. For each, name the transformation needed before calculating anything.
This trains selection rather than formula recall.
The transformation-chain drill
Give a raw data set and ask for change, rate, percentage, graph and conclusion in sequence.
At each stage, the learner labels the quantity and unit.
The reverse-data drill
Provide a rate and time interval and ask for total change; provide percentage and total and ask for the count.
Reverse transformations deepen understanding of the relationship among quantities.
The compare-two-transformations drill
Calculate both absolute and relative change, or both total output and normalised output.
Then explain which comparison answers the stated scientific aim.
The data-check drill
Insert one deliberate denominator or unit error into a worked solution. The learner uses units, reasonableness and raw data to locate it.
This builds checking into data handling.
The advanced standard
An advanced learner does not treat a table as a collection of numbers waiting for every available formula.
They identify the scientific question, choose the transformation that creates the required quantity, preserve units and baseline, and interpret the result within the evidence boundary.
Final perspective
Data transformation is scientific translation.
Raw readings become changes, rates, ratios, percentages, means and graphs only when those forms reveal the relationship the question asks about. Preserve the path from measurement to derived quantity so the conclusion can always be traced back to evidence.
One final data rule
Every derived value should be traceable back to raw evidence. If a rate, mean, percentage or graph point cannot be reconstructed from the recorded measurements and stated operations, the transformation chain has become opaque.
Keep that trace visible in important questions. It makes checking faster, prevents units from drifting and ensures the final scientific conclusion still belongs to the measurements that produced it.
The final transformation check
Before leaving a data question, state the transformed quantity in words: temperature decrease, average rate, percentage of total, output per gram or graph gradient. If you can name it precisely, the unit and denominator are easier to verify.
This one sentence keeps the calculation attached to scientific meaning and prevents a correct number from being used to answer the wrong data question.
