Statistics is not the art of decorating numbers with graphs. It is the discipline of organising data so that useful patterns become visible without changing what the data actually say. A good representation reveals structure. A poor one hides it. A misleading one can make a true dataset appear to support a false impression.
This Secondary 2 Mathematics Learning Guide develops data representation, measures of central tendency, grouped information and statistical interpretation. It also teaches students to inspect axes, scales, class intervals, categories and denominators before trusting a visual conclusion.
Secondary Mathematics Hub: S1–S4 Capability Map · Secondary 2 Learning Guide, Batch 3, Guide 3. Companion guides cover trigonometry, mensuration, and probability.
Course boundary. The current MOE G2/G3 Mathematics syllabus includes analysis and interpretation of statistical representations, measures such as mean, mode and median, grouped-data mean, and explaining why a statistical diagram may lead to misinterpretation. Exact sequencing differs by school. This guide focuses on those core ideas and clearly labels broader interpretation where appropriate.
Navigate: Start with the question · Representations · Mean, median, mode · Grouped data · Misleading graphs · Comparing datasets · Practice and answers · Teaching and transfer.
1. Statistical work begins with what the data represent
A list of numbers without context is incomplete. Are the numbers marks, ages, travel times, heights, temperatures or counts? What units are used? How many observations are present? Were the data measured, counted or categorised?
The same numerical pattern can carry different meanings in different contexts. A mean of 12 could be 12 minutes, 12 dollars, 12 students or 12 goals. Interpretation starts by naming the variable and unit.
Worked example 1: count before calculating
Suppose the data are 5, 7, 7, 8, 10, 13. There are six observations. Their total is 50. The mean is 50/6 ≈ 8.33. The median is the average of the third and fourth values, (7 + 8)/2 = 7.5. The mode is 7.
Three different measures describe the same dataset differently. None is automatically the best summary for every purpose.
2. Different representations reveal different features
A frequency table makes counts explicit. A dot diagram preserves individual observations while showing clustering. A stem-and-leaf diagram preserves exact values and order. A bar chart compares category frequencies. A histogram groups continuous numerical data into intervals and represents frequency across those intervals.
The representation should match the variable and the question. A bar chart is appropriate for categories such as transport mode. A histogram is appropriate for continuous intervals such as heights or travel times. The two may look similar but encode different structures.
Worked example 2: frequency table from raw data
Data: 2, 3, 3, 4, 4, 4, 5, 5, 6. A frequency table records value 2 with frequency 1, 3 with 2, 4 with 3, 5 with 2 and 6 with 1. Total frequency = 9, matching the number of observations.
Always check the total frequency against the raw-data count. If they disagree, an observation was lost or double-counted during organisation.
Worked example 3: stem-and-leaf structure
For values 21, 24, 24, 27, 31, 33 and 38, a stem-and-leaf diagram may use stems 2 and 3. The leaves for stem 2 are 1, 4, 4, 7; for stem 3 they are 1, 3, 8. A key such as 2|4 = 24 is essential.
Without the key, the same diagram might represent 2.4, 24 or 240 depending on context. Statistical notation needs declared scale just as coordinate graphs need labelled axes.
Worked example 4: dot diagram interpretation
A dot diagram with many observations near 6 and only one observation at 14 shows clustering near 6 with a high isolated value. Before computing anything, the diagram already suggests that the mean may be pulled upward more than the median.
A good statistical reader asks what the shape of the data suggests before reducing the dataset to one number.
3. Mean, median and mode answer different summary questions
The mean uses every numerical observation. The median is the middle value after ordering. The mode is the most frequent value or category. Because their definitions differ, they respond differently to unusual values.
Worked example 5: effect of an extreme value
Dataset A: 5, 6, 6, 7, 8. Mean = 6.4 and median = 6. Replace 8 with 28 to form Dataset B: 5, 6, 6, 7, 28. Mean becomes 10.4 while median remains 6.
The mean is sensitive because every value contributes to the total. The median depends on order position and is resistant to one extreme value when the middle position does not change.
Worked example 6: mean from a frequency table
Suppose score 1 occurs twice, score 2 occurs three times and score 4 occurs once. Total score = 1(2) + 2(3) + 4(1) = 12. Total frequency = 6. Mean = 12/6 = 2.
The multiplication value × frequency reconstructs the contribution of each repeated value without rewriting all six observations.
Choosing a useful measure
If a distribution contains a strong extreme value, the median may better describe a typical middle position. If total contribution matters, the mean may be more useful. If the most common category matters, the mode may be the only meaningful choice.
For a categorical variable such as favourite transport mode, calculating a mean is meaningless. The variable type constrains which statistics are valid.
4. Grouping compresses data and loses exact detail
When data are grouped into intervals such as 0–9, 10–19 and 20–29, the exact values inside each group are no longer known from the table. This makes displays and broad summaries easier but introduces approximation when estimating the mean.
Worked example 7: estimated mean from grouped data
Suppose travel times are grouped as 0–9 minutes: frequency 2; 10–19: frequency 5; 20–29: frequency 3. Use class midpoints 4.5, 14.5 and 24.5 as representative values.
Estimated total = 4.5(2) + 14.5(5) + 24.5(3) = 9 + 72.5 + 73.5 = 155. Total frequency = 10. Estimated mean = 15.5 minutes.
It is an estimate because the actual observations may not equal the midpoints. The grouped table has discarded exact within-class positions.
Class boundaries and intervals must not overlap ambiguously
For continuous data, classes such as 0 ≤ t < 10 and 10 ≤ t < 20 make membership unambiguous. A time of exactly 10 minutes belongs to the second class under those definitions.
If a table labels classes only as 0–10 and 10–20 without explaining the convention, the boundary value 10 appears to belong to both. Good statistical representation eliminates that ambiguity.
Histogram versus bar chart
Bars in a categorical bar chart are usually separated because categories are distinct. Histogram bars touch because the intervals form a continuous scale. If class widths differ, the relationship between bar dimensions and frequency requires additional care; follow the method taught in the current course.
5. A graph can be numerically correct and visually misleading
Misleading graphs often exploit scale, truncation, unequal intervals, selective time windows, area distortion or missing context. The numbers printed may all be true while the visual emphasis encourages an exaggerated conclusion.
Worked example 8: truncated vertical axis
Suppose two values are 98 and 102. On an axis from 0 to 110, they appear close. On an axis from 97 to 103, the 4-unit difference fills most of the graph height and appears dramatic.
A truncated axis is not automatically dishonest. It may be useful for showing small changes. But the reader must notice the starting value and avoid interpreting visual height as proportional to the full magnitude.
Worked example 9: unequal interval spacing
A time-series graph places 2020, 2021, 2022 and 2026 at equal horizontal gaps. The final gap represents four years while the earlier gaps represent one year. A line joining the points may visually imply a rate of change that does not match elapsed time.
Axes are measurement systems. Equal physical distances should represent equal numerical intervals unless a special scale is explicitly communicated.
Worked example 10: pictogram area distortion
If one icon is twice as tall and twice as wide as another, its area is four times as large. If the data value is only double, scaling the whole picture in two dimensions visually exaggerates the difference.
When pictograms encode value through repeated equal-size symbols, comparison is clearer. If symbol dimensions themselves vary, inspect which dimension is intended to represent the data.
Worked example 11: percentages without denominators
Group A reports 80% success from 10 trials: 8 successes. Group B reports 75% from 200 trials: 150 successes. The percentages compare rates, but the evidence base is very different in size.
A percentage is not a complete statistical story. Ask the denominator, sample size and collection method when the conclusion depends on reliability.
6. Read a statistical diagram with a five-question audit
- What variable is being shown?
- What do the axes, units and categories mean?
- What is the scale, and does it start at zero?
- Are intervals equal and categories comparable?
- What information is missing that would affect interpretation?
This audit prevents the learner from treating a visual impression as evidence before checking how the picture encodes the data.
7. Comparing datasets requires more than one statistic
Two datasets can have the same mean but very different distributions. Consider A: 5, 5, 5, 5, 5 and B: 1, 3, 5, 7, 9. Both have mean 5, but B varies much more.
At this level, even if formal spread measures are not yet central, the range and the visual distribution can reveal differences that the mean hides.
Worked example 12: same median, different shape
Dataset C: 3, 4, 5, 6, 7 has median 5. Dataset D: 0, 1, 5, 9, 10 also has median 5. The same middle value does not imply similar consistency or clustering.
Therefore a statistical comparison should identify what is similar and what remains different rather than relying on one summary number.
Context decides which difference matters
If comparing delivery times, a group with a slightly higher mean but much more consistent times may be preferred in one situation. In another, the lowest possible time may matter more. Mathematics summarises; the decision criterion comes from the context.
8. Common errors show different failures
- Median found before sorting: repair the definition of positional centre.
- Mean denominator is wrong: repair observation count or total frequency.
- Mode confused with largest value: repair frequency meaning.
- Bar chart used as histogram without interval reasoning: repair variable type.
- Grouped mean reported as exact: repair information-loss awareness.
- Truncated axis ignored: repair scale reading.
- Percentage compared without sample size: repair denominator awareness.
- Graph copied without labels or units: repair representation ownership.
9. Mixed practice: calculate, represent, question the picture
Questions 1–6. For data 4, 5, 5, 6, 8, 14: 1. Find the mean. 2. Find the median. 3. Find the mode. 4. Which measure is most affected by 14? 5. Replace 14 with 8 and recalculate the mean. 6. Explain why the median changes less than the mean.
Questions 7–12. 7. Scores 1, 2, 3 have frequencies 2, 5, 3. Find the mean. 8. State the total frequency. 9. A stem-and-leaf key is missing. Explain why interpretation is incomplete. 10. A graph compares values 49 and 51 using a vertical axis from 48 to 52. Why might the visual difference look exaggerated? 11. A timeline puts 2021, 2022 and 2026 at equal gaps. Identify the issue. 12. A pictogram doubles both height and width to show a doubled value. Why is this visually misleading?
Questions 13–18. 13. Grouped data have classes 0–9 frequency 3 and 10–19 frequency 5. Estimate the mean using midpoints. 14. Explain why the result is estimated. 15. Compare datasets A = 4,4,4,4,4 and B = 0,2,4,6,8. What measure is the same? 16. Which set has greater spread? 17. Group X has 90% success from 20 trials; Group Y has 85% from 400 trials. State one reason the percentages alone do not tell the whole story. 18. Give one situation where a truncated axis may be useful without being deceptive.
Explained answers: questions 1–6
1. Total = 42, so mean = 7. 2. Median = (5 + 6)/2 = 5.5. 3. Mode = 5. 4. Mean. 5. New total = 36, mean = 6. 6. The median depends on the central ordered positions, while the mean uses the numerical size of every observation.
Explained answers: questions 7–12
7. Total score = 1(2) + 2(5) + 3(3) = 21; total frequency 10; mean 2.1. 8. 10. 9. The place value or scale of stems and leaves is unknown. 10. A narrow truncated axis makes a 2-unit difference occupy a large fraction of the graph height.
11. Equal physical spacing represents unequal time intervals. 12. Doubling both dimensions makes the icon area four times as large, exaggerating a twofold data change.
Explained answers: questions 13–18
13. Midpoints 4.5 and 14.5: estimated total = 13.5 + 72.5 = 86; frequency 8; estimated mean = 10.75. 14. Exact values inside each interval are unknown and are represented by midpoints.
15. Both have mean 4 and median 4. 16. Dataset B. 17. Sample sizes differ greatly; the percentage does not show the denominator. 18. To display small but meaningful changes clearly, provided the truncated scale is visible and interpretation remains honest.
10. Teaching sequence: data meaning before calculation
Begin with one small dataset and ask students to organise it in three forms: ordered list, frequency table and dot diagram. Ask what each form makes easier to see. Then calculate mean, median and mode and discuss why the summaries differ.
Next introduce an extreme value and predict which statistic will move most before calculating. Then show two graphs of the same data with different axes. The learner should identify how the visual story changes even though the numbers do not.
Questions parents and tutors can ask
What does each observation represent? How many observations are there? Did you sort before finding the median? Why is this representation suitable? Does the axis begin at zero? Are the intervals equal? Is the mean exact or estimated? What information would you want before accepting the graph’s conclusion?
11. The transfer test: two truthful graphs, two different impressions
Suppose a quantity rises from 200 to 210. A graph with vertical axis 0–250 makes the change look modest. A graph with axis 198–212 makes the same change look steep. Both can plot the values accurately.
The correct conclusion is not that one graph must be forbidden. It is that statistical reading requires awareness of scale. A 5% increase remains a 5% increase regardless of how dramatic the line appears on the page.
This is statistical judgement: separate the data relationship from the visual rhetoric used to display it.
Name the variable. Check the denominator. Choose a representation that fits the data. Use summary statistics for the question they actually answer. Audit the scale before trusting the picture.
Continue to Probability, Sample Spaces and Combined Events · Return to the Secondary Mathematics Hub.
