← AS & A Level Mathematics

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.
Two Datasets with Equal Mean (30) but Different Variation Batter A (Consistent): Mean = 30 · Range = 12 Batter B (Erratic): Mean = 30 · Range = 64
Dot plot comparing two datasets with identical central location but radically different dispersion.

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%.
Anatomy of a Box-and-Whisker Diagram Min Q1 (25%) Q2 (Median) Q3 (75%) Max IQR = Q3 − Q1 (Middle 50%)
Box-and-whisker plot structure illustrating the five-figure summary and the central IQR box.

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.
Visualizing Distribution Skewness Positive skew: Long right tail (Q3 − Q2 > Q2 − Q1) Symmetrical: Balanced halves (Q3 − Q2 ≈ Q2 − Q1) Negative skew: Long left tail (Q2 − Q1 > Q3 − Q2)
Comparative box plots showing the geometric signature of positive, symmetric, and negative skew.

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.
Cumulative Frequency S-Curve & Percentile Lookup Variable x (e.g. Mass or Time) Cumulative frequency (cf) 0 n/4 (Q1) n/2 (Q2) 3n/4 (Q3) n (100%) Q1 Q2 (Median) Q3
Locating quartiles and percentiles on a cumulative frequency ogive by projecting horizontally from cf positions to the curve and down to the horizontal axis.

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.

  1. The 10 values are already arranged in ascending order. n = 10.
  2. Range = Maximum − Minimum = 83 − 18 = 65.
  3. Median Q2 is the average of the 5th and 6th values: (42 + 47)/2 = 44.5.
  4. Lower half = {18, 29, 39, 41, 42}. Lower quartile Q1 is the median of lower half = 39.
  5. Upper half = {47, 51, 64, 73, 83}. Upper quartile Q3 is the median of upper half = 64.
  6. Interquartile range IQR = Q3 − Q1 = 64 − 39 = 25.
  7. Upper outlier threshold = Q3 + 1.5 × IQR = 64 + 1.5(25) = 64 + 37.5 = 101.5.
  8. Since 83 ≤ 101.5, 83 is not an outlier.

✏️ Try it: Range & IQR

  1. 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.

  2. 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.

  3. 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.
Geometric Meaning of Variance: Average Area of Deviation Squares Mean μ = 20 Area = (−6)²
= 36 x₁=14 4²=16 x₂=24 Area = 10² = 100 x₃=30 Variance σ² = Mean of areas = (36 + 16 + 100)/3 ≈ 50.7
Variance geometrically represents the average area of squares built upon the distances from each data point to the mean.

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).
1. Compute sample meanμ = (Σx) / n (or Σfx / Σf for frequency tables)2. Compute sum of squaresCalculate Σx² (or Σfx² for frequency tables)3. Apply computational formulaVar(x) = (Σx² / n) − μ² (exact fractional arithmetic)4. Take positive square rootStandard deviation σ = √Var(x) with identical data units
Computational workflow for evaluating variance and standard deviation from summary totals.

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 =

Comparison: Standard Deviation vs Interquartile Range Standard Deviation (σ) ✔ Takes account of EVERY data point. ✔ Direct algebraic links to normal distribution. ✖ Heavily distorted by extreme outliers. ✖ Best suited for symmetric distributions. Interquartile Range (IQR) ✔ Highly resistant to extreme outliers. ✔ Best paired with the median for skewed data. ✖ Ignores outer 50% of the dataset completely. ✖ Less mathematically flexible in formal proofs.
Key strengths, trade-offs, and recommended application contexts for SD versus IQR.

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.

  1. Count the observations: n = 5.
  2. Calculate the sum of values: Σx = 3 + 5 + 8 + 12 + 17 = 45.
  3. Calculate the mean: μ = Σx / n = 45 / 5 = 9.
  4. Calculate the sum of squares: Σx² = 3² + 5² + 8² + 12² + 17² = 9 + 25 + 64 + 144 + 289 = 531.
  5. Apply computational formula: Var(x) = (Σx² / n) − μ² = (531 / 5) − 9² = 106.2 − 81 = 25.2.
  6. Take the positive square root: σ = √25.2 ≈ 5.02 (3 s.f.).
  7. 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.

  1. Total frequency: Σf = 4 + 12 + 9 + 5 = 30.
  2. Calculate Σfx: (0×4) + (1×12) + (2×9) + (3×5) = 0 + 12 + 18 + 15 = 45.
  3. Calculate mean: μ = Σfx / Σf = 45 / 30 = 1.5.
  4. Calculate Σfx²: (0²×4) + (1²×12) + (2²×9) + (3²×5) = 0 + 12 + 36 + 45 = 93.
  5. Calculate variance: Var(x) = (Σfx² / Σf) − μ² = (93 / 30) − 1.5² = 3.1 − 2.25 = 0.85.
  6. Standard deviation: σ = √0.85 ≈ 0.922 (3 s.f.).

✏️ Try it: Variance & SD

  1. 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.

  2. 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.

  3. 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.
Navigating Summary Statistics & Reconstructed Totals Summary Stats Sample size: n Mean μ and SD σ Σx = nμ Σx² = n(σ² + μ²) Reconstructed Totals Total sum: Σx Sum of squares: Σx²
Converting freely between descriptive statistics (n, μ, σ) and fundamental algebraic totals (n, Σx, Σx²).

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

Why Var(Combined) ≠ Average of Variances: The Two Components 1. Within-Group Variance Spread within Group X: Var(x) Spread within Group Y: Var(y) Measures internal scatter around
each group's own mean. + 2. Between-Group Variance Separation of means: |μ_x − μ_y| Difference from combined mean If μ_x ≠ μ_y, combined spread
expands substantially!
Combined variance accounts for both the dispersion inside each group and the distance separating the two group centers.

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.

  1. Total count: N = 10 + 15 = 25.
  2. Pooled sum: Σ(x + y) = 1650 + 2370 = 4020.
  3. Pooled sum of squares: Σ(x² + y²) = 275 490 + 377 835 = 653 325.
  4. Combined mean: μ_comb = 4020 / 25 = 160.8 cm.
  5. Combined variance: Var_comb = (653 325 / 25) − 160.8² = 26 133 − 25 856.64 = 276.36 cm².
  6. 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.

  1. Original group (n = 28): Σx = 28 × 172.6 = 4832.8 cm.
  2. 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.
  3. 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.
  4. New mean (n = 27): μ_new = 4671.0 / 27 = 173.0 cm.
  5. New variance: Var_new = (808 549.38 / 27) − (173.0)² = 29 946.273 − 29 929.0 = 17.273 cm².
  6. New standard deviation: σ_new = √17.273 ≈ 4.16 cm (3 s.f.).
  7. 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

  1. 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.

  2. 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!).

  3. 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.
Effect of Shift Transformation: y = x − 10 Original x: Distance = 160 Shifted y: Distance = 160 (UNCHANGED) Translating data points leaves all measures of dispersion invariant!
Shifting data translates the location of the distribution without altering the distance between any pair of points.

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).
Effect of Scaling: y = 2x (Stretching Spread) Data x: Gap = 80 Data y = 2x: Gap = 160 (Doubled!) Spread doubles: SD(2x) = 2 SD(x) · Variance quadruples: Var(2x) = 4 Var(x)
Multiplying by constant a stretches all point distances by |a|, scaling standard deviation by |a| and variance by a².

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
Coded variable uu = (x − b) / aFind coded mean & Varμ_u = Σu / n, Var(u) = Σu²/n − μ_u²Uncode the meanμ_x = a μ_u + b (shifts & scales)Uncode the spreadVar(x) = a² Var(u), σ_x = |a| σ_u
Coding and uncoding workflow for analyzing large or decimal datasets.

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.

  1. Define the coded variable: u = x − 10. Sample size n = 8.
  2. Coded totals: Σu = 48 and Σu² = 1490.
  3. Calculate coded mean: μu = Σu / n = 48 / 8 = 6.
  4. Calculate coded variance: Var(u) = (Σu² / n) − μu² = (1490 / 8) − 6² = 186.25 − 36 = 150.25.
  5. Uncode the mean: since u = x − 10 ⇒ x = u + 10, we have μx = μu + 10 = 6 + 10 = 16.
  6. Uncode the spread: shifting by 10 does not change spread, so Var(x) = Var(u) = 150.25.
  7. 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.

  1. Identify linear parameters: a = 1.8, b = 32.
  2. Calculate mean Fahrenheit: μF = 1.8μC + 32 = 1.8(12.5) + 32 = 22.5 + 32 = 54.5°F.
  3. Calculate variance Fahrenheit: Var(F) = a² Var(C) = (1.8)² × 20.25 = 3.24 × 20.25 = 65.61 (°F)².
  4. Calculate standard deviation Fahrenheit: σF = |a|σC = 1.8 × √20.25 = 1.8 × 4.5 = 8.1°F.
  5. Check: √Var(F) = √65.61 = 8.1. Consistent!

✏️ Try it: Linear Coding

  1. 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.

  2. 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.

  3. 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

Algebraic Proof of the Computational Variance Formula 1. Definition: Var(x) = (1/n) Σ(x − μ)² 2. Expand brackets: (x − μ)² = x² − 2μx + μ² 3. Sum totals: Σ(x − μ)² = Σx² − 2μ(Σx) + nμ² = Σx² − 2μ(nμ) + nμ² = Σx² − nμ² 4. Divide by n: Var(x) = (Σx² / n) − μ² Q.E.D. Identical logic applies to grouped data replacing n with Σf and x with class midpoints.
Formal proof showing how the computational formula naturally emerges from expanding the squared deviations.

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

    1. 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).

    2. 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.

    3. 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:

    📄 Download PDF Assignment (with space for working)

    1. Find the range of the numbers −4.5, 2.1, −1.8, 6.7, 0, 3.4.
    2. Seven numbers are 4, 7, 9, 12, 15, 18, 22. Find the interquartile range (IQR).
    3. In a box plot, Q1 = 24 and Q3 = 40. Which value is classified as an outlier?
    4. For a dataset, Σx = 120, Σx² = 1840, and n = 10. Find the variance.
    5. For a sample of 25 values, Σ(x − 8)² = 225. Find the standard deviation if the mean is 8.
    6. If every value in a dataset is increased by 5, what happens to the standard deviation?
    7. The variable y is coded as y = 3x − 4. If Var(x) = 16, find the standard deviation of y.
    8. The variable u is defined by u = (x − 20)/5. If Var(u) = 4, find Var(x).
    9. Given n = 20, Σx = 100, and Var(x) = 15, find the value of Σx².
    10. Set A has 10 values with mean 12, Set B has 10 values with mean 18. Find the combined mean.
    11. Why is the interquartile range (IQR) preferred over the range when outliers are present?
    12. A box plot has Q1 = 15, Q2 = 28, and Q3 = 35. What is the skewness of this distribution?

    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.