Representation and summary of data
S1 Topic 2, spec point by spec point (Representation and summary of data). Each spec point has the method, worked examples written the way the mark scheme wants them, how the marks go and the mistakes that cost the most marks.
Spec2.1Histograms, stem and leaf diagrams and box plots
Use histograms, stem and leaf diagrams (including back-to-back) and box plots to compare distributions.
Histograms: area is frequency
In a histogram the area of each bar represents the frequency, which is why unequal class widths work:
\[\text{frequency density} = \frac{\text{frequency}}{\text{class width}}\]
When a question gives the size of one bar in cm, find the scale "1 cm² represents ... items" from that bar, then use it for the others. The width scale comes from the class widths.
Stem and leaf and box plots
- Stem and leaf: shows every value, in order, with a key (e.g. 4 | 3 means 43). Back-to-back diagrams compare two groups.
- Box plot: lowest value, \(Q_1\), median, \(Q_3\), highest value. Outliers are drawn as crosses and the whiskers go to the most extreme values that aren't outliers.
Comparing distributions
Always compare one measure of location (median or mean) and one measure of spread (IQR or standard deviation), in context. For example: "The median time for group A is higher, so group A generally took longer; the IQR for A is smaller, so their times were more consistent."
bar dimensions
The bar for \(32 \leqslant x < 36\) (frequency 20) is 2.5 cm wide and 7.5 cm tall. The class \(30 \leqslant x < 32\) has frequency 16. Find the width and height of its bar.
Width: a class of width 4 is 2.5 cm, so width 2 is 1.25 cm.
Area scale: \(2.5 \times 7.5 = 18.75\) cm² represents 20, so 16 is represented by \(\frac{16}{20} \times 18.75 = 15\) cm².
Height \(= \frac{15}{1.25} = 12\) cm.
- B1 for the width, M1 for using area \(\propto\) frequency (or frequency density), A1 for the height (WST01 May 2024 Q3a, WST01A Jan 2026 Q5a). Scaling the height without using areas loses the M mark.
- Box plots: M1 for a box with two whiskers, A1ft for the quartiles, A1 for the whisker ends, A1 for outliers marked (WST01 Oct 2021 Q3d).
Often asked as: bar sizes on a histogram; estimate a frequency from a histogram (area); draw or read a box plot; compare two distributions.
Plotting frequency instead of frequency density.
Whiskers drawn to an outlier.
Comparisons with no context, or two measures of location and none of spread.
Spec2.2Measures of location: mean, median, mode
Find and interpret the mean, median and mode for discrete, continuous, grouped or ungrouped data, and use coding.
The measures
- Mean \(\bar{x} = \dfrac{\sum x}{n}\), or \(\dfrac{\sum fx}{\sum f}\) from a frequency table. For grouped data use the midpoints: it's an estimate.
- Median: the middle value. For grouped continuous data, use linear interpolation: find the class containing the \(\frac{n}{2}\)th value and assume values are spread evenly through it.
- Mode: the most common value (modal class for grouped data).
Interpolation: median \(= L + \dfrac{\frac{n}{2} - F}{f} \times w\), where \(L\) is the lower class boundary, \(F\) the cumulative frequency before the class, \(f\) the class frequency and \(w\) the width.
Which to use?
The mean uses every value but is pulled by outliers and skew. The median is resistant to outliers, so it's better for skewed data.
Coding
If \(y = \dfrac{x - a}{b}\), then \(\bar{x} = a + b\bar{y}\). (Subtracting \(a\) shifts the mean; dividing by \(b\) scales it.)
interpolation
225 speeds are grouped; the class \(25 \leqslant x < 30\) has frequency 90, with 54 values below 25. Estimate the median.
Median position \(\frac{225}{2} = 112.5\). Median \(= 25 + \dfrac{112.5 - 54}{90} \times 5 = 28.3\).
effect of a correction
A bean recorded as 0.95 g was actually 1.4 g. Without recalculating, what happens to the mean?
The total increases, so the mean increases.
- Interpolation: M1 for a correct expression using \(\frac{n}{2}\), A1 for awrt the value (WST01 Jun 2025 Q5c).
- Effects of changes: B1 for each effect, dB1 for a complete reason (WST01A Jan 2026 Q5).
- "Show the mean is 3.5": products shown and divided by \(n\) (WST03 Jun 2017 Q4a).
Often asked as: median and quartiles by interpolation; mean from a frequency table; read the median from a stem and leaf diagram; effect of correcting or adding a value.
Using class widths instead of midpoints for a grouped mean.
Using \(\frac{n + 1}{2}\) for grouped continuous data.
Spec2.3Measures of dispersion
Find and interpret variance, standard deviation, range and interpercentile ranges, using interpolation where needed.
The measures
- Range = largest − smallest. Interquartile range IQR \(= Q_3 - Q_1\) (the middle 50%). Interpercentile range, e.g. \(P_{90} - P_{10}\).
- Variance (S1 uses divisor \(n\)): \(\sigma^2 = \dfrac{\sum x^2}{n} - \bar{x}^2 = \dfrac{\sum fx^2}{\sum f} - \bar{x}^2\), "mean of the squares minus square of the mean".
- Standard deviation \(\sigma = \sqrt{\text{variance}}\), in the same units as the data.
Coding
If \(y = \dfrac{x - a}{b}\), then \(\sigma_x = b\,\sigma_y\). Adding or subtracting a constant doesn't change the spread.
Working backwards
From the mean and variance you can recover the sums: \(\sum x = n\bar{x}\) and \(\sum x^2 = n(\sigma^2 + \bar{x}^2)\).
Effects of adding a value
Adding a value equal to the mean leaves the mean unchanged but decreases the standard deviation (same total squared deviation, divided by a bigger \(n\)).
recover the sums
10 values have mean 163.5 and variance 148.05. Find \(\sum x^2\).
\(\sum x^2 = 10(148.05 + 163.5^2) = 268\,803\).
lower quartile by interpolation
98 birth weights: 16 are below 2.5 kg and the class 2.5–3.0 kg has 24. Estimate \(Q_1\).
Position \(\frac{98}{4} = 24.5\). \(Q_1 = 2.5 + \dfrac{24.5 - 16}{24} \times 0.5 = 2.68\) kg.
- B1* for the correct expression with an intermediate step for a printed answer, M1 A1 for a found sum (WST01 Jun 2026 Q5a).
- Interpolation: M1 using position \(\frac{n}{4}\) (using 25 instead of 24.5 is M0), A1 (WST01 Jan 2020 Q4b).
- Effects: B1 mean unchanged, B1 standard deviation decreases, ddB1 for a valid reason (WST01 May 2024 Q3e).
Often asked as: mean and standard deviation from a frequency table or from sums; quartiles by interpolation; effect of extra values; coded data.
\(\frac{\sum x^2}{n} - \bar{x}\) (forgetting to square the mean).
Changing the standard deviation when a constant is added.
Spec2.4Skewness and outliers
Describe skewness, and identify outliers using a rule given in the question.
Skewness
| Shape | Quartiles | Mean, median, mode |
|---|---|---|
| Symmetrical | \(Q_3 - Q_2 = Q_2 - Q_1\) | mean \(=\) median \(=\) mode |
| Positive skew (long tail to the right) | \(Q_3 - Q_2 > Q_2 - Q_1\) | mean \(>\) median \(>\) mode |
| Negative skew (long tail to the left) | \(Q_3 - Q_2 < Q_2 - Q_1\) | mean \(<\) median \(<\) mode |
Questions may also give a coefficient like \(\frac{3(\text{mean} - \text{median})}{\text{standard deviation}}\): its sign gives the skew.
Outliers
An outlier is an unusually extreme value. The rule is always given, typically: below \(Q_1 - 1.5(Q_3 - Q_1)\) or above \(Q_3 + 1.5(Q_3 - Q_1)\).
When asked whether to remove an outlier, think about the context: a genuine but unusual value should stay; an error should be removed.
an outlier
\(Q_1 = 28\) and \(Q_3 = 48\). Show 90 is an outlier using \(Q_3 + 1.5(Q_3 - Q_1)\).
\(48 + 1.5(20) = 78\). Since \(90 > 78\), 90 is an outlier.
skewness
\(Q_1 = 20\), \(Q_2 = 26\), \(Q_3 = 30\). \(Q_2 - Q_1 = 6 > Q_3 - Q_2 = 4\), so the data are negatively skewed.
- B1 for a quartile, M1 for the fence calculation, A1* for both quartiles correct and the comparison (WST01 Jan 2024 Q4a).
- Skewness: M1 for comparing \(Q_2 - Q_1\) with \(Q_3 - Q_2\) (with values), A1ft for the conclusion (WST01 Oct 2020 Q4c).
- Is a Normal model suitable? B1 for "skewed, not symmetrical", dB1 for a second reason such as "the data are discrete" (WST01A Jan 2026 Q2d).
Often asked as: find the outlier limits and list outliers; describe skew with a reason; comment on whether a Normal model fits.
Saying "positive skew" with no supporting calculation.
Inventing an outlier rule instead of using the one given.