Measures of Central Tendency & Dispersion: Concepts, Calculations & Applications

TAKEAWAYS:

• Central tendency (mean, median, mode) tells you the "typical" value, but dispersion (range, variance, standard deviation) shows how spread out the data is—always report both to avoid misleading conclusions.
• Mean is sensitive to outliers; median is robust; mode is for categorical data or most frequent values—choose based on data type and skewness.
• Grouped data requires midpoints and frequency weights; ungrouped data uses raw values—always check if data is grouped before calculating.
• Variance measures squared deviation; standard deviation is its square root—use standard deviation for interpretation (it’s in the same units as the data).
• Coefficient of variation (CV) compares dispersion across datasets with different units—critical for benchmarking (e.g., comparing stock volatility).
• Exam focus: 30% of questions test calculation (raw/grouped data), 40% test conceptual choice (why mean vs. median?), and 30% test real-world application (e.g., interpreting bus wait times or loan interest rates).

1. Measures of Central Tendency: The "Typical" Value

Central tendency summarizes a dataset with a single value. The choice depends on data type (nominal, ordinal, interval, ratio) and skewness (symmetry).

1.1 Definitions and When to Use Each

classDiagram
    class Measure {
        +Name: String
        +DataType: String
        +Skewness: String
        +Formula: String
        +UseCase: String
    }
    class Mean {
        +Name: "Mean (Average)"
        +DataType: "Interval/Ratio"
        +Skewness: "Symmetric or mild skew"
        +Formula: "ΣX / N"
        +UseCase: "Most common for quantitative data"
    }
    class Median {
        +Name: "Median"
        +DataType: "Ordinal/Interval/Ratio"
        +Skewness: "Skewed data (robust to outliers)"
        +Formula: "(n+1)/2th value (ungrouped)"
        +UseCase: "Income, house prices (outliers distort mean)"
    }
    class Mode {
        +Name: "Mode"
        +DataType: "Nominal/Ordinal"
        +Skewness: "Any"
        +Formula: "Most frequent value"
        +UseCase: "Categorical data (e.g., bus routes used)"
    }
    Measure <|-- Mean
    Measure <|-- Median
    Measure <|-- Mode

Key Idea:

• Mean = Total sum / Number of values (sensitive to outliers).
• Median = Middle value (50th percentile; robust to outliers).
• Mode = Most frequent value (can be multiple; used for categories).

Visual: Data Skewness and Measure Choice

Worked Example 1: Choosing the Right Measure Problem: The following are the waiting times (in minutes) for a bus over 20 days: 15, 10, 2, 17, 5, 8, 3, 10, 12, 18, 4, 6, 9, 11, 7, 14, 1, 20, 13, 5 Step 1: Check for outliers. Observation: The data is right-skewed (one extreme value: 20 minutes). The mean will be higher than the median. Step 2: Calculate all three measures.

• Mean = $\frac{\sum X}{N}=\frac{183}{20}=9.15$ minutes.
• Median = Average of 10th and 11th values = $\frac{10+11}{2}=10.5$ minutes.
• Mode = 5 and 10 (bimodal).

Conclusion: The median (10.5) best represents the "typical" wait time because the mean is inflated by the 20-minute outlier.

1.2 Calculations for Grouped Data

When data is presented in classes (e.g., age groups), use midpoints and frequency weights.

Formula:

• Mean = $\frac{\sum (f\times m)}{\sum f}$ Where:
  • $f$ = frequency of the class
  • $m$ = midpoint of the class = $\frac{\text{lower limit}+\text{upper limit}}{2}$

Worked Example 2: Grouped Data Mean Problem: Calculate the mean age from the following table:

Step 1: Assign midpoints.

Step 2: Calculate mean. $$\text{Mean}=\frac{3275}{100}=32.75\text{ years}$$

2. Measures of Dispersion: How Spread Out the Data Is

Dispersion tells you how much the data varies. High dispersion = data is spread out; low dispersion = data is clustered.

2.1 Absolute vs. Relative Measures

Key Idea:

• Absolute measures (range, IQR, SD) depend on the data’s units.
• Relative measures (CV) are unitless and allow comparison (e.g., stock volatility).

Visual: Dispersion in Action

Worked Example 3: Variance and Standard Deviation Problem: Calculate the variance and standard deviation for the bus wait times from Worked Example 1. Step 1: Compute the mean ($\bar{X}=9.15$). Step 2: Calculate squared deviations from the mean. Step 3: Compute variance and SD. $$\text{Variance}=\frac{1020.75}{20}=51.0375$$ $$\text{SD}=\sqrt{51.0375}\approx 7.14\text{ minutes}$$

Interpretation: The typical wait time varies by ±7.14 minutes from the mean.

2.2 Coefficient of Variation (CV)

Formula: $$CV=\left(\frac{\text{SD}}{\text{Mean}}\right)\times 100\%$$ Worked Example 4: Comparing Dispersion Problem: Two companies, A and B, have the following monthly salaries (in thousands):

• Company A: Mean = 40, SD = 5
• Company B: Mean = 60, SD = 10

Which has more consistent salaries? Step 1: Calculate CV for both. $$C{V}_A=\left(\frac{5}{40}\right)\times 100=12.5\%$$ $$C{V}_B=\left(\frac{10}{60}\right)\times 100\approx 16.67\%$$ Conclusion: Company A has more consistent salaries (lower CV).

3. Real-World Applications in Nepal

3.1 eSewa and Khalti: Transaction Times

• Idea Used: Central tendency (median) and dispersion (IQR).
• How:
  • eSewa reports the "typical" transaction time as the median (robust to slow outliers).
  • The IQR shows how much times vary (e.g., "Most transactions take 5–10 seconds, but some take up to 20").
  • Why not mean? A few very slow transactions (e.g., due to network issues) would skew the average.

Visual: eSewa Transaction Times

3.2 Ncell and NTC: Customer Wait Times

• Idea Used: Mean vs. median for skewed data.
• How:
  • Ncell’s customer service reports the average wait time (mean) for calls, but the median is more reliable because some calls take much longer (e.g., technical issues).
  • Dispersion (SD) helps them identify high-variability periods (e.g., weekends vs. weekdays).

Worked Example 5: Ncell Call Center Problem: Ncell records call wait times (minutes) for 30 customers: 2, 3, 4, 5, 5, 6, 7, 8, 9, 10, 12, 15, 20, 25, 30, 3, 4, 5, 6, 7, 8, 9, 10, 12, 15, 20, 25, 30, 35, 40 Step 1: Check skewness. Observation: Right-skewed (long waits are outliers). Step 2: Calculate measures.

• Mean = 12.5 minutes (inflated by 35 and 40).
• Median = 10 minutes (better "typical" value).
• SD = 11.2 minutes (high dispersion).

Conclusion: Ncell should advertise the median (10 minutes) and work to reduce the SD (e.g., hire more agents during peak hours).

3.3 Daraz and Pathao: Order Fulfillment Times

• Idea Used: Standard deviation for performance benchmarking.
• How:
  • Daraz tracks the SD of delivery times to ensure consistency. A high SD means some orders take much longer than others (e.g., due to traffic or weather).
  • Pathao uses IQR to set "guaranteed delivery windows" (e.g., "90% of orders arrive within 15–30 minutes").

Worked Example 6: Pathao Delivery Times Problem: Pathao records delivery times (minutes) for 20 orders: 12, 15, 18, 20, 22, 25, 28, 30, 32, 35, 14, 16, 19, 21, 24, 26, 29, 31, 33, 40 Step 1: Compute five-number summary.

• Min = 12
• Q1 = 18th value = 19
• Median (Q2) = Average of 10th and 11th = 22.5
• Q3 = 29
• Max = 40 Step 2: Calculate IQR. $$IQR=Q3-Q1=29-19=10\text{ minutes}$$ Interpretation: Pathao can promise "Most deliveries take between 19 and 29 minutes" (Q1 to Q3), with only 25% of orders outside this range.

4. Common Pitfalls and Exam Traps

1. Ignoring Data Type:

   • ❌ Using mean for nominal data (e.g., bus routes: "Route 1", "Route 2").
   • ✅ Use mode for categories.

2. Forgetting Midpoints for Grouped Data:

   • ❌ Calculating mean without converting classes to midpoints.
   • ✅ Always use $m=\frac{\text{lower}+\text{upper}}{2}$.

3. Mixing Population and Sample Variance:

   • Population variance: Divide by $N$.
   • Sample variance: Divide by $n-1$ (for unbiased estimate).

4. Overlooking Skewness:

   • ❌ Reporting mean for highly skewed data.
   • ✅ Use median for skewed distributions.

5. Units in Dispersion:

   • ❌ Comparing SD of salary (₹) and height (cm) directly.
   • ✅ Use CV for unitless comparison.

Exam Tip: How to Score Full Marks

1. Always State the Formula:

   • Write the formula before plugging in numbers. Example:

     "Mean = $\frac{\sum X}{N}$. Here, $\sum X=183$ and $N=20$, so Mean = $\frac{183}{20}=9.15$."

2. Justify Your Choice:

   • If asked to pick a measure, explain why. Example:

     "The median is more appropriate here because the data is right-skewed (presence of extreme values like 20 and 40), and the median is robust to outliers."

3. Show All Steps for Grouped Data:

   • Exam markers deduct marks if you skip midpoint calculations. Always show a table like this:

     | Class | Midpoint (m) | Frequency (f) | f × m |
     |-------|--------------|---------------|-------|
     | ...   | ...          | ...           | ...   |
     

4. Interpret Results:

   • Don’t just compute; explain what it means. Example:

     "A standard deviation of 7.14 minutes means that most bus wait times fall within ±7 minutes of the mean (9.15 minutes), i.e., between 2 and 16 minutes."

5. Watch for Tricks:

   • Open-ended classes: For "50 and more", assume midpoint = 55 (or use upper limit if specified).
   • Sample vs. population: If the problem says "sample," use $n-1$ for variance.

Quick Revision Table

Final Worked Example: Comprehensive Problem

Problem: The following table shows the marks obtained by 130 students in a Statistics exam:

i) Compute the mean marks. ii) Compute the standard deviation. iii) Interpret the results.

Solution: i) Mean Calculation:

$$\text{Mean}=\frac{7290}{130}\approx 56.08\text{ marks}$$

ii) Standard Deviation: Step 1: Compute $(X-\bar{X}{)}^2$ for each class. Step 2: Compute variance and SD. $$\text{Variance}=\frac{209700}{130}\approx 1613.08$$ $$\text{SD}=\sqrt{1613.08}\approx 40.16\text{ marks}$$

iii) Interpretation:

• The average mark is 56.08, but the SD of 40.16 shows high variability. This suggests:
  • Some students scored very low (<40), while others scored near perfect (80+).
  • The exam may have been difficult for some or had uneven question distribution.
  • Action: Check if the exam had easy and hard sections or if teaching was inconsistent.

Summary Checklist for Exams

Before submitting, ensure you’ve covered:

1. Defined all terms (mean, median, mode, range, SD, CV).
2. Calculated all required measures (show tables for grouped data).
3. Justified your choice of measure (e.g., "median for skewed data").
4. Interpreted results in real-world terms (e.g., "high SD means inconsistent performance").
5. Avoided common mistakes (e.g., using mean for nominal data, forgetting midpoints).