Measures of Dispersion

Unit III · Lecture 19

BDS113 — Introduction to Statistics

Coefficient of Variation

Relative Dispersion · Comparing Variability · Formula & Applications · Properties & Limitations

Unit III · Lecture 19 of 40 · B.Sc. Data Science, 1st Year, Semester II

Measures of Dispersion

Unit III · Lecture 19

The Journey — Lecture 19

🤔

Why Relative Measures?
Limitations of absolute measures · The need for comparison

📐

Coefficient of Variation — Definition & Formula
CV formula · Interpretation · Step-by-step calculation

✏️

Worked Examples & Applications
Comparing two datasets · Real-world scenarios · Interactive tools

⚖️

Properties, Comparisons & Quiz
CV vs SD · When to use CV · Limitations · Practice quiz

Measures of Dispersion

Unit III · Lecture 19

🤔 Why Do We Need Relative Measures of Dispersion?

💡 The Problem: Comparing Apples and Oranges

Suppose you want to compare the consistency of two athletes:

Sprinter A (100m times in seconds): 10.2, 10.5, 10.3, 10.4, 10.6 → Mean = 10.4s, SD = 0.14s

Marathon Runner B (times in minutes): 125, 135, 130, 140, 130 → Mean = 132 min, SD = 5.10 min

Sprinter A has SD = 0.14 and Runner B has SD = 5.10. Does this mean Runner B is far more inconsistent? No! Their SDs are in completely different units (seconds vs minutes) and their means are vastly different. We cannot directly compare 0.14 seconds with 5.10 minutes.

We need a unit-free, relative measure that expresses variability as a proportion of the mean.

📖 Absolute vs Relative Measures

Absolute measures (Range, MAD, Variance, SD): Expressed in the same units as the data. Cannot be used to compare datasets with different units or vastly different means.

Relative measures (Coefficient of Variation, Coefficient of MAD): Expressed as ratios or percentages — unit-free. Allow meaningful comparison across different datasets, units, and scales.

Measures of Dispersion

Unit III · Lecture 19

📐 Coefficient of Variation — Definition & Formula

📖 Formal Definition

The Coefficient of Variation (CV) is a relative (unit-free) measure of dispersion that expresses the standard deviation as a percentage of the mean. It was introduced by Karl Pearson and is also known as the Coefficient of Standard Deviation (when expressed as a ratio, not percentage).

📖 Formula

CV = (σ / x̄) × 100%

Where:
  • σ = standard deviation of the dataset
  • x̄ = mean of the dataset
  • The result is expressed as a percentage (%)

Note: When expressed as a simple ratio (without × 100), it is called the Coefficient of Standard Deviation: CSD = σ / x̄

💡 Analogy: Error Rate

Think of CV as the "error rate" of your data relative to its average. If a factory produces bolts with mean diameter 10 cm and SD 0.1 cm, the CV = 1%. If another factory produces beams with mean diameter 100 cm and SD 1 cm, the CV is also 1%. Despite different SDs (0.1 vs 1), both factories have the same relative variability — and that's exactly what CV captures.

Measures of Dispersion

Unit III · Lecture 19

🔍 Interpreting the Coefficient of Variation

📖 How to Read CV

Lower CV → Data points are closer to the mean → More consistent, more homogeneous, more reliable.

Higher CV → Data points are more spread out from the mean → More variable, more heterogeneous, less reliable.

CV = 0% → All values are identical (no variability at all).

✅ When CV is Low (e.g., < 10%)

• Manufacturing quality is consistent
• Test scores are uniform
• Investment returns are stable
• Data is homogeneous

⚠️ When CV is High (e.g., > 30%)

• Manufacturing quality is inconsistent
• Test scores show wide variation
• Investment returns are volatile
• Data is heterogeneous

💬 Important: There is no universal "good" or "bad" CV threshold — it depends on the field. In精密 manufacturing, a CV of 2% might be unacceptable. In social sciences, a CV of 20% might be perfectly normal. Context matters.

Measures of Dispersion

Unit III · Lecture 19

🪜 Step-by-Step: Calculating CV

1 Find the Mean (x̄): x̄ = Σxᵢ / n (or Σfᵢxᵢ / N for grouped data)

2 Find the Standard Deviation (σ): Using the direct method, shortcut method, or step-deviation method.

3 Apply the CV Formula: CV = (σ / x̄) × 100%

4 Interpret: Compare CVs across datasets. The dataset with the lower CV is more consistent.

💡 Caution: CV is meaningful only when the mean is positive and non-zero. If the mean is zero or negative (e.g., temperature in °C, profit/loss), CV becomes undefined or misleading. In such cases, use absolute measures (SD, MAD) instead.

Measures of Dispersion

Unit III · Lecture 19

✏️ Example 1: Comparing Consistency of Two Students

📝 Problem

Two students scored the following marks in 5 tests (out of 100):

Student A: 70, 75, 80, 85, 90

Student B: 40, 60, 80, 100, 120

Who is more consistent in performance?

📖 Solution — Student A

x̄ₐ = (70+75+80+85+90)/5 = 400/5 = 80
Σ(xᵢ−x̄)² = (−10)²+(−5)²+0²+5²+10² = 100+25+0+25+100 = 250
σₐ² = 250/5 = 50 → σₐ = √50 = 7.07
CVₐ = (7.07/80) × 100 = 8.84%

📖 Solution — Student B

x̄_b = (40+60+80+100+120)/5 = 400/5 = 80
Σ(xᵢ−x̄)² = (−40)²+(−20)²+0²+20²+40² = 1600+400+0+400+1600 = 4000
σ_b² = 4000/5 = 800 → σ_b = √800 = 28.28
CV_b = (28.28/80) × 100 = 35.35%

Comparison Result

CV_A = 8.84% < CV_B = 35.35%

Student A is more consistent

Both students have the same mean (80), but Student A's scores are much closer to the mean. CV quantifies this difference.

Measures of Dispersion

Unit III · Lecture 19

✏️ Example 2: Comparing Different Units

📝 Problem

A company wants to compare the variability of two product lines:

Product A (weight in grams): Mean = 500g, SD = 10g

Product B (length in cm): Mean = 25cm, SD = 2cm

Which product line has more relative variability?

📖 Product A

CV_A = (10/500) × 100 = 2%

📖 Product B

CV_B = (2/25) × 100 = 8%

Comparison Result

CV_A = 2% < CV_B = 8%

Product B is more variable

Even though Product B has a smaller absolute SD (2 vs 10), its variability relative to its mean is higher. CV makes this comparison possible despite different units (grams vs cm).

Measures of Dispersion

Unit III · Lecture 19

✏️ Example 3: CV of Grouped Data

📝 Problem

The wages (in ₹) of workers in two factories are distributed as follows. Compare their variability using CV.

Factory A

Wages (₹)	fᵢ
100–150	5
150–200	12
200–250	18
250–300	10
300–350	5

Factory B

Wages (₹)	fᵢ
100–150	8
150–200	10
200–250	14
250–300	12
300–350	6

📖 Solution (Using Assumed Mean Method)

Factory A: x̄ = 222, σ = 55.14, CV_A = (55.14/222) × 100 = 24.84%
Factory B: x̄ = 218, σ = 68.26, CV_B = (68.26/218) × 100 = 31.31%

Conclusion: CV_B (31.31%) > CV_A (24.84%). Factory B has greater relative variability in wages. Factory A pays its workers more consistently.

💡 Note: Even though both factories have similar mean wages (₹222 vs ₹218), the CV reveals that Factory B's wage distribution is more spread out relative to its mean.

Measures of Dispersion

Unit III · Lecture 19

🎬 Interactive: CV Calculator & Comparer

Enter two datasets to calculate and compare their CVs side by side:

Dataset A (comma-separated):

Dataset B (comma-separated):

Measures of Dispersion

Unit III · Lecture 19

🎬 Visual: Consistency Meter

Click each dataset to see its CV and a visual consistency rating:

🎯

Precision Bolts

Mean=10cm, SD=0.1cm

📏

Ruler Factory

Mean=30cm, SD=1.5cm

🧪

Lab Chemicals

Mean=50ml, SD=5ml

📈

Stock Returns

Mean=12%, SD=8%

Measures of Dispersion

Unit III · Lecture 19

⚠️ When NOT to Use CV

📖 Important Restrictions

CV is a powerful tool, but it has specific conditions under which it is meaningful. Using CV inappropriately leads to misleading conclusions.

Mean is zero: If x̄ = 0, the denominator is zero, and CV is undefined (division by zero).
Mean is negative: If x̄ < 0 (e.g., losses, temperature below zero), CV produces a negative percentage, which is meaningless for comparison.
Interval scale data: For data measured on an interval scale (e.g., temperature in °C, calendar years), the zero point is arbitrary. CV is meaningful only for ratio scale data (where zero means "absence" of the quantity — weight, height, income).
Means are vastly different: When means differ by orders of magnitude, CV comparison may be misleading because the SD often scales with the mean.
Small means close to zero: Even if the mean is technically positive but very close to zero, CV can become extremely large and uninformative.

💬 Rule of Thumb: Use CV only when the data is on a ratio scale (true zero, meaningful ratios) and the mean is positive and reasonably far from zero. For interval data or data with negative values, stick to absolute measures (SD, MAD).

Measures of Dispersion

Unit III · Lecture 19

🔑 Properties of Coefficient of Variation

CV is unit-free: Since it is a ratio of SD to mean (both in the same units), the units cancel out. CV is expressed as a pure percentage.
CV is always non-negative: Since SD ≥ 0 and we consider |mean|, CV ≥ 0. It equals zero only when SD = 0 (all values identical).
CV allows comparison across different units: You can compare the variability of height (cm) with weight (kg) or income (₹) — something impossible with SD alone.
CV is independent of change of origin: Adding a constant to all values changes the mean but not the SD. However, since the mean changes, CV also changes. (This is a key difference from SD.)
CV is affected by change of scale: If y = c × x, then SD(y) = |c| × SD(x) and mean(y) = c × mean(x), so CV(y) = CV(x). CV is invariant under scaling.
Lower CV = more consistency: When comparing two or more datasets, the one with the lower CV is more consistent (less variable relative to its mean).
CV is sensitive to small means: A small mean amplifies the CV, even if the absolute SD is modest. This can be misleading.

Measures of Dispersion

Unit III · Lecture 19

⚖️ CV vs Absolute Measures — When to Use What

Scenario	Best Measure	Reason
Same units, similar means	SD or MAD	Direct comparison is valid
Different units	CV	Units cancel out; comparison becomes meaningful
Same units, very different means	CV	SD is misleading when means differ greatly
Data with negative values or zero mean	SD or MAD	CV is undefined or meaningless
Quick intuitive understanding	MAD	Simplest to compute and interpret
Advanced statistical analysis	SD / Variance	Algebraically tractable; used in tests and models
Comparing reliability / consistency	CV	Standardised measure; widely used in quality control

💡 Real-World Applications of CV

Quality Control: Comparing consistency of production across factories with different output levels.
Finance: Comparing risk (volatility) of investments with different average returns. Lower CV = better risk-adjusted return.
Medicine: Comparing variability of lab test results across different assays with different measurement scales.
Agriculture: Comparing yield variability across different crops (wheat in kg/hectare vs cotton in kg/hectare).
Education: Comparing score variability across subjects with different maximum marks.

Measures of Dispersion

Unit III · Lecture 19

✏️ Example 4: Investment Comparison (Finance)

📝 Problem

An investor is choosing between two stocks based on their annual returns over 5 years:

Stock X: Returns = 8%, 10%, 12%, 14%, 16%

Stock Y: Returns = 2%, 6%, 10%, 14%, 18%

Which stock offers more consistent returns?

📖 Solution

Stock X:
Mean = (8+10+12+14+16)/5 = 12%
SD = √[(16+4+0+4+16)/5] = √8 = 2.83%
CV_X = (2.83/12) × 100 = 23.58%

Stock Y:
Mean = (2+6+10+14+18)/5 = 10%
SD = √[(64+16+0+16+64)/5] = √32 = 5.66%
CV_Y = (5.66/10) × 100 = 56.57%

Investment Decision

CV_X = 23.58% < CV_Y = 56.57%

Stock X is more consistent

Stock X offers more predictable returns. A risk-averse investor would prefer Stock X. Note: Stock X also has a higher mean return (12% vs 10%), making it superior on both counts in this example.

Measures of Dispersion

Unit III · Lecture 19

🔑 Summary: Coefficient of Variation

CV = (σ / x̄) × 100% — a relative, unit-free measure of dispersion introduced by Karl Pearson.
Purpose: To compare the variability of two or more datasets that may have different units, scales, or means.
Interpretation: Lower CV → more consistent / homogeneous. Higher CV → more variable / heterogeneous.
CV is unit-free because the units of SD and mean cancel out in the ratio.
CV is invariant under scaling (multiplying all values by a constant does not change CV) but not invariant under origin shift (adding a constant changes the mean and hence the CV).
Restrictions: CV is meaningful only for ratio-scale data with a positive, non-zero mean. Do not use CV when the mean is zero, negative, or close to zero.
Applications: Quality control, finance (risk-return comparison), medicine, agriculture, education — anywhere relative variability needs to be compared.
CV vs SD: Use SD when comparing datasets with similar units and means. Use CV when units differ or means are very different.
Key formula reminder: Coefficient of SD = σ/x̄ (ratio). CV = (σ/x̄) × 100 (percentage).

Measures of Dispersion

Unit III · Lecture 19

🧠 Quick Quiz — Lecture 19

Q1: The Coefficient of Variation is defined as:

A) (σ / x̄)

B) (σ / x̄) × 100%

C) (x̄ / σ) × 100%

D) σ² / x̄

Q2: CV is useful when comparing datasets that have:

A) The same mean and same units

B) Different units or different means

C) Only negative values

D) A mean of zero

Q3: Dataset A has CV = 5% and Dataset B has CV = 15%. Which is more consistent?

A) Dataset A

B) Dataset B

C) Both are equally consistent

D) Cannot be determined

Q4: If all values in a dataset are multiplied by 5, the CV will:

A) Increase 5 times

B) Decrease to 1/5th

C) Remain the same

D) Increase by 5

Q5: CV is NOT meaningful when:

A) The data has large values

B) The mean is zero or negative

C) The data is on a ratio scale

D) The standard deviation is small

Q6: Two datasets have the same SD but different means. The dataset with the larger mean will have:

A) A smaller CV

B) A larger CV

C) The same CV

D) CV = 0

Q7: The mean salary of Company A is ₹50,000 with SD ₹5,000. The mean salary of Company B is ₹30,000 with SD ₹4,500. Which company has more relative variability in salaries?

A) Company A (CV = 10%)

B) Company B (CV = 15%)

C) Both have the same CV

D) Cannot be determined

Q8: Who introduced the Coefficient of Variation?

A) Ronald Fisher

B) Karl Pearson

C) John Tukey

D) Florence Nightingale

Measures of Dispersion

Unit III · Lecture 19

📝 Homework & Practice

📝 Practice Tasks

1. Calculate the CV for the following dataset: 25, 30, 35, 40, 45, 50

2. Two batsmen scored the following runs in 6 innings:
   Batsman A: 45, 50, 55, 60, 65, 70
   Batsman B: 20, 40, 50, 60, 80, 95
   (a) Calculate the mean and SD for each.
   (b) Calculate the CV for each.
   (c) Who is more consistent? Explain using CV.

3. The mean and SD of marks in Mathematics are 60 and 12 respectively. The mean and SD of marks in Physics are 50 and 10 respectively. Which subject shows more relative variability?

4. A factory produces bolts with mean diameter 5 cm and SD 0.05 cm. Another factory produces pipes with mean diameter 50 cm and SD 0.5 cm. Compare their relative variability using CV.

5. Explain why CV cannot be used when the mean temperature is −5°C.

6. If the CV of a dataset is 20% and the mean is 250, find the standard deviation.

7. Dataset A has mean 100, SD 10. If every value is increased by 50, what happens to the CV? If every value is multiplied by 2, what happens to the CV? Verify mathematically.

8. BONUS: A company claims its product has a CV of 0.5%. Is this good or bad? Under what circumstances would a very low CV be desirable? Give two real-world examples.

💡 Tip: When solving CV problems, always check: (1) Is the mean positive? (2) Is the data on a ratio scale? If either answer is "no," reconsider whether CV is the right measure.

Measures of Dispersion

Unit III · Lecture 19

Thank You!

Questions? Let's discuss!

Upcoming in Unit III

Lecture 20: Combined Variance & SD of Two Groups
Lecture 21: Lorenz Curve — Measuring Inequality Graphically
Lecture 22: Review & Problem-Solving Session

Unit III · Lecture 19 · BDS113 Introduction to Statistics · B.Sc. Data Science, Semester II