AS & A Level Mathematics · Probability & Statistics 1 · Chapter 3
Measures of variation
- A measure of central tendency alone gives an incomplete picture: two datasets can have identical means yet entirely different degrees of spread.
- You will master the range and the interquartile range (IQR), identifying why IQR provides a robust measure resistant to extreme outliers.
- You will construct and interpret box-and-whisker plots, evaluate skewness, and apply the 1.5 × IQR fence rule to classify outliers.
- You will calculate variance (σ²) and standard deviation (σ) for raw, ungrouped, and grouped continuous data using both definitional and computational formulae.
- You will manipulate summary totals (n, Σx, Σx²) to combine datasets and uncover why combined variance is never the simple average of individual variances.
- You will unlock linear coding y = ax + b, proving that shifting data leaves spread untouched while scaling multiplies standard deviation by |a| and variance by a².
Learning objectives
- Calculate the range and understand its sensitivity to extreme values.
- Find the lower quartile (Q1), median (Q2), upper quartile (Q3), and interquartile range (IQR).
- Estimate medians, quartiles, and percentiles from cumulative frequency polygons and curves.
- Construct and interpret box-and-whisker plots and assess positive, negative, or symmetrical skewness.
- Apply the mathematical definition of outliers: values beyond Q1 − 1.5×IQR or Q3 + 1.5×IQR.
Book exercises: Exercise 3A.
1. The Need for Measures of Variation
- Central tendency vs variation: knowing the average (mean, median, or mode) locates the center, but tells nothing about how widely dispersed the values are.
- Range definition: Range = Maximum value − Minimum value.
- Limitation: relies solely on the two extreme values and is heavily distorted by even a single outlier.
2. Quartiles and the Interquartile Range (IQR)
- Quartiles: divide an ordered dataset into four equal segments, each containing 25% of the observations.
- Q1 (Lower Quartile): separates the bottom 25% from the top 75%.
- Q2 (Median): the middle value (50th percentile).
- Q3 (Upper Quartile): separates the bottom 75% from the top 25%.
- Interquartile range formula: IQR = Q3 − Q1. Measures the spread of the middle 50% of the data.
- Advantage: completely unaffected by extreme outliers in the top or bottom 25%.
3. Assessing Skewness from Box Plots
- Positive skew (skewed right): Q3 − Q2 > Q2 − Q1, right whisker noticeably longer than left whisker.
- Symmetrical: Q3 − Q2 ≈ Q2 − Q1, median is central within the box, whiskers approximately equal.
- Negative skew (skewed left): Q2 − Q1 > Q3 − Q2, left whisker noticeably longer than right whisker.
4. Identifying Outliers: The 1.5 × IQR Rule
- Mathematical definition: an observation x is an outlier if:
- x < Q1 − 1.5 × IQR (Lower boundary / lower fence)
- x > Q3 + 1.5 × IQR (Upper boundary / upper fence)
- In box plots, outliers are plotted as individual isolated crosses (×) or dots beyond the fences; whiskers terminate at the nearest non-outlier value.
Worked Example 1 · Discrete Quartiles & Outliers
A sample of 10 student quiz marks is: 18, 29, 39, 41, 42, 47, 51, 64, 73, 83. Find the range, the interquartile range, and determine whether 83 is an outlier.
- The 10 values are already arranged in ascending order. n = 10.
- Range = Maximum − Minimum = 83 − 18 = 65.
- Median Q2 is the average of the 5th and 6th values: (42 + 47)/2 = 44.5.
- Lower half = {18, 29, 39, 41, 42}. Lower quartile Q1 is the median of lower half = 39.
- Upper half = {47, 51, 64, 73, 83}. Upper quartile Q3 is the median of upper half = 64.
- Interquartile range IQR = Q3 − Q1 = 64 − 39 = 25.
- Upper outlier threshold = Q3 + 1.5 × IQR = 64 + 1.5(25) = 64 + 37.5 = 101.5.
- Since 83 ≤ 101.5, 83 is not an outlier.
✏️ Try it: Range & IQR
-
Find the range and IQR of: 7, 13, 21, 2, 37, 28, 17, 11, 2.
Solution
Step 1: Sort 9 values: 2, 2, 7, 11, 13, 17, 21, 28, 37.
Step 2: Range = 37 − 2 = 35.
Step 3: Median Q2 is 5th value = 13.
Step 4: Lower half = {2, 2, 7, 11} ⇒ Q1 = (2 + 7)/2 = 4.5. Upper half = {17, 21, 28, 37} ⇒ Q3 = (21 + 28)/2 = 24.5.
Step 5: IQR = 24.5 − 4.5 = 20.
Answer: Range = 35, IQR = 20. -
A dataset has Q1 = 15 and Q3 = 35. Find the range of non-outlier values if Min = 0 and Max = 68.
Solution
Step 1: IQR = 35 − 15 = 20.
Step 2: Lower fence = Q1 − 1.5(IQR) = 15 − 30 = −15.
Step 3: Upper fence = Q3 + 1.5(IQR) = 35 + 30 = 65.
Step 4: Min = 0 > −15 (not an outlier); Max = 68 > 65 (68 is an outlier).
Step 5: Answer: Values between 0 and 65 inclusive. -
In a cumulative frequency graph of 120 cars, the 30th car is at 42 km/h, the 60th at 50 km/h, and the 90th at 61 km/h. State the median, IQR, and skewness.
Solution
Step 1: Q1 = 42 km/h, Q2 = 50 km/h, Q3 = 61 km/h.
Step 2: IQR = Q3 − Q1 = 61 − 42 = 19 km/h.
Step 3: Q2 − Q1 = 50 − 42 = 8; Q3 − Q2 = 61 − 50 = 11.
Step 4: Since Q3 − Q2 > Q2 − Q1 (11 > 8), the distribution has positive skew.
Answer: Median = 50 km/h, IQR = 19 km/h, Positive skew.
Learning objectives
- Explain why mean deviation is zero and why squared deviations are used to define variance.
- Master the definitional formula Var(x) = Σ(x − μ)² / n.
- Master the efficient computational formula Var(x) = (Σx² / n) − μ².
- Calculate variance and standard deviation for discrete frequency tables and grouped continuous distributions.
- Critically evaluate the advantages and disadvantages of standard deviation versus interquartile range.
Book exercises: Exercise 3B.
1. Why Square the Deviations?
- The deviation of a data point x from the mean μ is (x − μ).
- The sum of all raw deviations is identically zero: Σ(x − μ) = Σx − nμ = nμ − nμ = 0. Positive and negative deviations cancel!
- Taking absolute values Σ|x − μ| / n gives the mean absolute deviation, but absolute values are algebraically intractable in calculus and probability models.
- Squaring deviations makes every deviation non-negative: (x − μ)² ≥ 0, penalizes large deviations proportionally, and allows rigorous algebraic manipulation.
= 36
2. Definitional vs Computational Formula
- Definitional formula:
Var(x) = σ² =
Ideal when the mean μ is a convenient integer or simple terminating decimal. - Computational formula (recommended):
Var(x) = − μ² = − ()²
"The mean of the squares minus the square of the mean." - Standard deviation:
σ = √Var(x) = √( − μ² )
Standard deviation shares the exact same physical units as the original observations (e.g. cm, kg, seconds).
3. Grouped Data & Frequency Tables
- When values x appear with frequencies f:
μ = , Var(x) = − μ², σ = √Var(x)
- For continuous grouped intervals (e.g. 20–30 kg), x is taken as the midpoint of each class interval:
Midpoint x =
Worked Example 2 · Raw Data Variance & Standard Deviation
Calculate the exact mean, variance, and standard deviation (to 3 s.f.) of: 3, 5, 8, 12, 17.
- Count the observations: n = 5.
- Calculate the sum of values: Σx = 3 + 5 + 8 + 12 + 17 = 45.
- Calculate the mean: μ = Σx / n = 45 / 5 = 9.
- Calculate the sum of squares: Σx² = 3² + 5² + 8² + 12² + 17² = 9 + 25 + 64 + 144 + 289 = 531.
- Apply computational formula: Var(x) = (Σx² / n) − μ² = (531 / 5) − 9² = 106.2 − 81 = 25.2.
- Take the positive square root: σ = √25.2 ≈ 5.02 (3 s.f.).
- Check via definitional formula: Σ(x − 9)² = (−6)² + (−4)² + (−1)² + 3² + 8² = 36 + 16 + 1 + 9 + 64 = 126. Var = 126 / 5 = 25.2. Verified!
Worked Example 3 · Frequency Table Standard Deviation
The numbers of defective components produced per batch are: 0 (f = 4), 1 (f = 12), 2 (f = 9), 3 (f = 5). Calculate the standard deviation.
- Total frequency: Σf = 4 + 12 + 9 + 5 = 30.
- Calculate Σfx: (0×4) + (1×12) + (2×9) + (3×5) = 0 + 12 + 18 + 15 = 45.
- Calculate mean: μ = Σfx / Σf = 45 / 30 = 1.5.
- Calculate Σfx²: (0²×4) + (1²×12) + (2²×9) + (3²×5) = 0 + 12 + 36 + 45 = 93.
- Calculate variance: Var(x) = (Σfx² / Σf) − μ² = (93 / 30) − 1.5² = 3.1 − 2.25 = 0.85.
- Standard deviation: σ = √0.85 ≈ 0.922 (3 s.f.).
✏️ Try it: Variance & SD
-
Find the variance and standard deviation of: 10, 14, 15, 17, 24.
Solution
Step 1: n = 5, Σx = 10 + 14 + 15 + 17 + 24 = 80 ⇒ μ = 80/5 = 16.
Step 2: Σx² = 100 + 196 + 225 + 289 + 576 = 1386.
Step 3: Var(x) = (1386 / 5) − 16² = 277.2 − 256 = 21.2.
Step 4: σ = √21.2 ≈ 4.60.
Answer: Var = 21.2, SD = 4.60. -
For 20 numbers, Σx = 140 and Σx² = 1260. Find the standard deviation.
Solution
Step 1: μ = 140 / 20 = 7.
Step 2: Var(x) = (1260 / 20) − 7² = 63 − 49 = 14.
Step 3: σ = √14 ≈ 3.74.
Answer: SD = √14 ≈ 3.74. -
A grouped distribution has Σf = 50, Σfx = 1200, and Σfx² = 30 000. Calculate the standard deviation.
Solution
Step 1: μ = 1200 / 50 = 24.
Step 2: Var(x) = (30 000 / 50) − 24² = 600 − 576 = 24.
Step 3: σ = √24 = 2√6 ≈ 4.90.
Answer: SD ≈ 4.90.
Learning objectives
- Reconstruct sums of squares Σx² from given sample size n, mean μ, and variance or standard deviation.
- Calculate the combined mean and combined standard deviation when two distinct datasets are merged.
- Understand and explain why combined variance is NOT the average of the two component variances.
- Update summary statistics when an observation is added, corrected, or removed.
Book exercises: Exercise 3C.
1. Reconstructing Σx² from Summary Parameters
- By rearranging the computational formula:
Var(x) = − μ² ⇒ Σx² = n[Var(x) + μ²] = n[σ² + ()²]
- This algebraic rearrangement is essential when raw observations are withheld and only summary totals (n, μ, σ) are reported.
2. Combining Two Sets of Data
- Suppose Set X has nx items with totals Σx and Σx², while Set Y has ny items with totals Σy and Σy².
- When merged into a single pooled dataset of size N = nx + ny:
Σcomb = Σx + Σy, Σcomb² = Σx² + Σy²
- The combined mean is:
μcomb =
- The combined variance is:
Varcomb = − (μcomb)²
each group's own mean.
expands substantially!
Worked Example 4 · Combining Boys' & Girls' Heights
The heights, x cm, of 10 boys have Σx = 1650 and Σx² = 275 490. The heights, y cm, of 15 girls have Σy = 2370 and Σy² = 377 835. Calculate the combined mean and standard deviation of all 25 children.
- Total count: N = 10 + 15 = 25.
- Pooled sum: Σ(x + y) = 1650 + 2370 = 4020.
- Pooled sum of squares: Σ(x² + y²) = 275 490 + 377 835 = 653 325.
- Combined mean: μ_comb = 4020 / 25 = 160.8 cm.
- Combined variance: Var_comb = (653 325 / 25) − 160.8² = 26 133 − 25 856.64 = 276.36 cm².
- Combined standard deviation: σ_comb = √276.36 ≈ 16.6 cm (3 s.f.).
Worked Example 5 · Correcting Data When an Outlier Leaves
A group of 28 people has mean height 172.6 cm and standard deviation 4.58 cm. A person of height 161.8 cm leaves. Find the new mean and standard deviation of the remaining 27 people.
- Original group (n = 28): Σx = 28 × 172.6 = 4832.8 cm.
- Original sum of squares: Σx² = 28 × [(4.58)² + (172.6)²] = 28 × [20.9764 + 29 790.76] = 28 × 29 811.7364 = 834 728.62 cm².
- Subtract the departing member (x = 161.8):
New Σx = 4832.8 − 161.8 = 4671.0 cm.
New Σx² = 834 728.62 − (161.8)² = 834 728.62 − 26 179.24 = 808 549.38 cm². - New mean (n = 27): μ_new = 4671.0 / 27 = 173.0 cm.
- New variance: Var_new = (808 549.38 / 27) − (173.0)² = 29 946.273 − 29 929.0 = 17.273 cm².
- New standard deviation: σ_new = √17.273 ≈ 4.16 cm (3 s.f.).
- Notice: removing a low extreme value caused the mean to rise (172.6 → 173.0) and spread to decrease (4.58 → 4.16). Checked!
✏️ Try it: Totals & Combined Sets
-
Given n = 25, Σx = 275, and Var(x) = 7, find Σx².
Solution
Step 1: μ = 275 / 25 = 11.
Step 2: Var(x) = Σx²/25 − 11² ⇒ 7 = Σx²/25 − 121.
Step 3: Σx²/25 = 128 ⇒ Σx² = 25 × 128 = 3200.
Answer: Σx² = 3200. -
Set 1: 10 values with mean 12, Var = 4. Set 2: 10 values with mean 18, Var = 4. Find the combined variance.
Solution
Step 1: Set 1: Σx = 120, Σx² = 10(4 + 144) = 1480.
Step 2: Set 2: Σy = 180, Σy² = 10(4 + 324) = 3280.
Step 3: N = 20, Σcomb = 300 ⇒ μ_comb = 15.
Step 4: Σcomb² = 1480 + 3280 = 4760.
Step 5: Var_comb = 4760/20 − 15² = 238 − 225 = 13.
Answer: Combined variance = 13 (much greater than 4 due to the gap between means!). -
Five values have mean 50 and SD 10. A 6th value of 50 is added. State the new mean and determine whether variance increases, decreases, or stays the same.
Solution
Step 1: Adding a value exactly equal to the mean keeps the mean unchanged: New mean = 50.
Step 2: The added value contributes (50 − 50)² = 0 to the sum of squared deviations.
Step 3: Σ(x − 50)² was 5 × 100 = 500. Now it is still 500, but divided by 6 instead of 5.
Step 4: New variance = 500 / 6 = 83.33 < 100.
Answer: Mean = 50; Variance decreases to 83.3.
Learning objectives
- Analyze the effect of shifting data y = x − b on mean, range, IQR, variance, and standard deviation.
- Analyze the effect of scaling data y = ax on mean, range, IQR, variance, and standard deviation.
- Apply general linear coding y = ax + b and u = (x − b)/a to simplify complex data sets.
- Recover true original statistics (μx, Var(x), σx) from coded totals Σ(x − b) and Σ(x − b)².
Book exercises: Exercises 3D & 3E.
1. Effect of Addition/Subtraction: y = x − b (Shifting)
- When a constant b is subtracted from or added to every observation:
- Mean shifts: μy = μx − b.
- Spread is UNCHANGED: Range(y) = Range(x), IQR(y) = IQR(x), Var(y) = Var(x), σy = σx.
- Geometric intuition: Sliding all points along the number line preserves every relative gap between them.
2. Effect of Multiplication: y = ax (Scaling)
- When every observation is multiplied by a non-zero scale factor a:
- Mean scales: μy = aμx.
- Range and IQR scale: Range(y) = |a| Range(x), IQR(y) = |a| IQR(x).
- Standard deviation scales: σy = |a| σx. (Note the absolute value: SD is strictly non-negative!)
- Variance scales quadratically: Var(y) = a² Var(x).
3. General Linear Coding: y = ax + b
| Measure | Original Data x | Coded Data y = ax + b | Mathematical Reason |
|---|---|---|---|
| Mean | μx | μy = aμx + b | Affected by both scaling and translation |
| Range | Rx | Ry = |a| Rx | Translation cancels: (axmax + b) − (axmin + b) = a(xmax − xmin) |
| IQR | IQRx | IQRy = |a| IQRx | Constant b subtracts out: Q₃(y) − Q₁(y) = a(Q₃ − Q₁) |
| Variance | Var(x) = σx² | Var(y) = a² Var(x) | Deviations are squared: [(ax + b) − (aμ + b)]² = a²(x − μ)² |
| Std Deviation | σx | σy = |a| σx | σy = √Var(y) = √(a²σx²) = |a|σx |
Worked Example 6 · Uncoding Summary Totals
Eight values of x are summarized by Σ(x − 10) = 48 and Σ(x − 10)² = 1490. Find the mean and standard deviation of x.
- Define the coded variable: u = x − 10. Sample size n = 8.
- Coded totals: Σu = 48 and Σu² = 1490.
- Calculate coded mean: μu = Σu / n = 48 / 8 = 6.
- Calculate coded variance: Var(u) = (Σu² / n) − μu² = (1490 / 8) − 6² = 186.25 − 36 = 150.25.
- Uncode the mean: since u = x − 10 ⇒ x = u + 10, we have μx = μu + 10 = 6 + 10 = 16.
- Uncode the spread: shifting by 10 does not change spread, so Var(x) = Var(u) = 150.25.
- Standard deviation: σx = √150.25 ≈ 12.3 (3 s.f.).
Worked Example 7 · Temperature Scale Conversion
Daily temperatures recorded in Celsius have mean 12.5°C and variance 20.25 (°C)². Convert these to Fahrenheit using F = 1.8C + 32.
- Identify linear parameters: a = 1.8, b = 32.
- Calculate mean Fahrenheit: μF = 1.8μC + 32 = 1.8(12.5) + 32 = 22.5 + 32 = 54.5°F.
- Calculate variance Fahrenheit: Var(F) = a² Var(C) = (1.8)² × 20.25 = 3.24 × 20.25 = 65.61 (°F)².
- Calculate standard deviation Fahrenheit: σF = |a|σC = 1.8 × √20.25 = 1.8 × 4.5 = 8.1°F.
- Check: √Var(F) = √65.61 = 8.1. Consistent!
✏️ Try it: Linear Coding
-
The standard deviation of x is 5. Find the standard deviation of y = 3x − 7.
Solution
Step 1: Identify scale factor: a = 3.
Step 2: Translation by −7 has zero effect on spread.
Step 3: σy = |a|σx = 3 × 5 = 15.
Answer: SD(y) = 15. -
The variable u is defined by u = (x − 50)/10. If Var(u) = 2.4, find Var(x) and σx.
Solution
Step 1: Rearrange: x = 10u + 50, so scale factor a = 10.
Step 2: Var(x) = a² Var(u) = 10² × 2.4 = 100 × 2.4 = 240.
Step 3: σx = √240 ≈ 15.5.
Answer: Var(x) = 240, SD(x) ≈ 15.5. -
Twenty readings of y have Σ(y − 5) = 130 and Σ(y − 5)² = 890. Find the standard deviation of y.
Solution
Step 1: Let u = y − 5. Mean μu = 130 / 20 = 6.5.
Step 2: Var(u) = (890 / 20) − 6.5² = 44.5 − 42.25 = 2.25.
Step 3: Since shifting leaves SD unchanged: σy = σu = √2.25 = 1.5.
Answer: SD(y) = 1.5.
Learning objectives
- Follow the rigorous algebraic proof demonstrating the equivalence of the two variance formulae.
- Use an interactive live simulation to see how data spread directly changes variance and standard deviation.
- Utilize an interactive solver to calculate variance, standard deviation, coded totals, and combined datasets with full step-by-step working.
- Master decision criteria for choosing between Range, IQR, and Standard Deviation in Cambridge 9709 examinations.
Book appendix & review: Appendix to Section 3.3.
1. Algebraic Proof: Equivalence of the Two Variance Formulae
- Let x₁, x₂, …, xn be n observations with sample mean μ = (1/n)Σx.
- Starting with the definitional formula:
Var(x) = Σ(x − μ)² =
- Expand the squared bracket: (x − μ)² = x² − 2xμ + μ².
- Distribute the summation:
Σ(x − μ)² = Σx² − 2μΣx + Σμ²
- Notice that μ is a constant: Σμ² = nμ², and Σx = nμ, so 2μΣx = 2μ(nμ) = 2nμ².
- Substitute these back:
Σ(x − μ)² = Σx² − 2nμ² + nμ² = Σx² − nμ²
- Dividing through by n:
Var(x) = = − μ²
2. Interactive Simulation: Spread & Standard Deviation Explorer
Drag the slider to adjust the spread (scale) of the data points or shift them along the number line. Watch how distances (x − μ) and squared distances (x − μ)² update live!
⚡ Live Dispersion & Variance Explorer
3. Interactive Tool: Measures of Variation Solver
🧮 Comprehensive Variation & Coding Calculator
Select calculation mode, enter values, and view complete step-by-step working.
✏️ Try it: Proofs & Choosing Measures
-
Explain why the standard deviation of the numbers 1004, 1007, 1009, 1012 is exactly equal to the standard deviation of 4, 7, 9, 12.
Solution
Step 1: The transformation is y = x − 1000.
Step 2: Subtracting a constant translates every value equally along the axis without altering the distance between any pair of points.
Step 3: Since variance measures squared deviations (x − μ)² and (x − 1000 − (μ − 1000))² = (x − μ)², variance and standard deviation are strictly identical.
Answer: Translational invariance of spread: SD(x − b) = SD(x). -
A salary distribution in a tech startup has a few multi-millionaire founders and 100 junior engineers. State which measure of central tendency and which measure of variation are most appropriate, giving reasons.
Solution
Step 1: The distribution is heavily positively skewed with extreme outliers in the upper tail.
Step 2: The mean and standard deviation are strongly distorted upward by the founders' huge salaries.
Step 3: The median represents the typical salary, and the interquartile range (IQR) represents the variation of the middle 50% without distortion by outliers.
Answer: Median and Interquartile Range. -
Given that Var(x) = 9, find the variance of y = 5 − 2x.
Solution
Step 1: Linear form: y = ax + b with a = −2 and b = 5.
Step 2: Var(y) = a² Var(x) = (−2)² × 9 = 4 × 9 = 36.
Answer: Var(y) = 36.
Checklist of Learning & Understanding
- Range: Max − Min. Highly sensitive to extreme values.
- Interquartile Range: IQR = Q₃ − Q₁. Resistant to outliers, measuring the spread of the middle 50%.
- Box-and-whisker plot: Displays the five-figure summary (Min, Q₁, Q₂, Q₃, Max).
- Outlier definition: Any observation x is an outlier if x < Q1 − 1.5×IQR or x > Q3 + 1.5×IQR.
- Skewness:
- Positive (right): Q3 − Q2 > Q2 − Q1, longer right whisker.
- Symmetrical: Q3 − Q2 ≈ Q2 − Q1, equal whiskers.
- Negative (left): Q2 − Q1 > Q3 − Q2, longer left whisker.
- Variance & Standard Deviation:
Var(x) = − μ², σ = √Var(x)
For frequency tables:Var(x) = − μ², μ =
- Linear coding y = ax + b:
μy = aμx + b, Var(y) = a² Var(x), σy = |a|σx, IQRy = |a|IQRx
- Combined datasets: Pooled size N = nx + ny, pooled sum Σ = Σx + Σy, pooled sum of squares Σ² = Σx² + Σy². Combined variance accounts for both within-group and between-group spread.
Practice Questions & Full Printable Assignment
Download the complete printable assignment with every question from Exercises 3A–3E, End-of-chapter review 3, Cross-topic review 1, and Check yourself questions:
Questions adapted from Cambridge International AS & A Level Mathematics: Probability & Statistics 1 Coursebook (Cambridge University Press, Dean Chalmers). Chapter 3: Measures of variation, Exercises 3A to 3E, End-of-chapter review exercise 3 and Cross-topic review exercise 1.