Measurement Scales & Data Types: Classification, Examples & Applications

TAKEAWAYS:

• Measurement scales (nominal, ordinal, interval, ratio) determine how data can be analyzed—nominal (labels only) to ratio (true zero + arithmetic).
• Data types split into qualitative (descriptive) and quantitative (numerical), with subcategories discrete (countable) and continuous (measurable).
• eSewa uses ordinal scales for user ratings (1–5 stars) and ratio for transaction amounts (Rs. 0–∞).
• Khalti applies interval scales for time (e.g., 12:00 PM vs. 12:30 PM) and discrete data for transaction counts.
• Worked examples show how to classify real datasets (e.g., Daraz order quantities vs. customer feedback).
• Exam tip: Always label scales/data types in answers and justify choices (e.g., "Temperature in °C is interval because it has equal intervals but no true zero").

1. Measurement Scales: The Four Levels of Data

Measurement scales define how data is categorized, ordered, or quantified. They determine which statistical operations are valid. Below is a hierarchy of scales from least to most informative:

graph LR
    A["Nominal"] --> B["Ordinal"]
    B --> C["Interval"]
    C --> D["Ratio"]
    A -->|"Labels only"| E["Categories"]
    B -->|"Order + labels"| F["Rankings"]
    C -->|"Equal intervals"| G["Temperature (°C), IQ"]
    D -->|"True zero + arithmetic"| H["Weight (kg), Age (years)"]

Key Definitions & Examples

A pyramid showing nominal (bottom) to ratio (top) with icons for each scale type.

Worked Example 1: Classifying Data from eSewa

Dataset: eSewa user feedback survey responses:

• Q1: "How satisfied are you with our service?" (Options: Very Dissatisfied, Dissatisfied, Neutral, Satisfied, Very Satisfied).
• Q2: "How much did you pay for your last transaction?" (Rs. 500, Rs. 1200, Rs. 3500, etc.).
• Q3: "What is your age group?" (18–25, 26–35, 36–45, 45+).

Solution:

1. Q1: Ordinal (ordered categories but no equal intervals between "Dissatisfied" and "Neutral").
2. Q2: Ratio (true zero, arithmetic valid: Rs. 1200 is 2.4× Rs. 500).
3. Q3: Ordinal (age groups are ordered but intervals are unequal).

Interpretation: The mode (most frequent) is "Satisfied," but we cannot calculate the mean satisfaction score because the scale is ordinal.

2. Data Types: Qualitative vs. Quantitative

Data is further classified into qualitative (descriptive) and quantitative (numeric), with subcategories:

graph TD
    A["Data Types"] --> B["Qualitative"]
    A --> C["Quantitative"]
    B --> D["Nominal/Ordinal"]
    C --> E["Discrete"]
    C --> F["Continuous"]
    D -->|"Text/Labels"| G["Gender, Blood Group"]
    E -->|"Countable"| H["Number of Daraz orders, Ncell calls"]
    F -->|"Measurable"| I["Height, Temperature, Traffic speed"]

Definitions & Examples

A Venn diagram showing qualitative (left) and quantitative (right) with subtypes and icons (e.g., a bar chart for discrete, a line graph for continuous).

Worked Example 2: Classifying Data from NTC

Dataset: NTC’s monthly electricity consumption (in kWh) for 10 households:

• Household A: 250, 260, 275, 280, 290
• Household B: Low, Medium, High, Low, Medium
• Household C: "High", "Very High", "High", "Medium", "Low"

Solution:

1. Household A: Continuous quantitative (measurable, infinite possible values between 250 and 290).
2. Household B: Ordinal qualitative (ordered categories but no numeric values).
3. Household C: Nominal qualitative (labels only, no order implied by "High" vs. "Very High").

Key Insight: Only Household A can have its mean consumption calculated (271 kWh). Households B and C require non-parametric methods.

3. Real-World Applications in Nepalese Tech

Example 1: eSewa (Ordinal + Ratio Scales)

• Ordinal: User ratings (1–5 stars) for service quality.
  • Why? "5 stars" > "3 stars," but the difference between "5" and "4" isn’t numerically meaningful.
• Ratio: Transaction amounts (Rs. 500, Rs. 2000).
  • Why? Rs. 0 is valid (no transaction), and Rs. 2000 is 4× Rs. 500.

Example 2: Khalti (Interval + Discrete Data)

• Interval: Time of transactions (e.g., 10:30 AM vs. 11:00 AM).
  • Why? 30-minute intervals are equal, but "10:30 AM" isn’t twice "5:15 AM."
• Discrete: Number of transactions per user (0, 1, 2, ...).
  • Why? You can’t have a fraction of a transaction.

Example 3: Daraz (Nominal + Ratio Data)

• Nominal: Product categories (Electronics, Grocery, Fashion).
  • Why? No order or arithmetic applies.
• Ratio: Order quantities (1 kg, 5 kg, 10 kg of rice).
  • Why? 10 kg is 10× 1 kg, and 0 kg is valid (no order).

4. Common Mistakes & How to Avoid Them

A flowchart showing "Is the data nominal/ordinal/interval/ratio?" with arrows to correct operations.

5. Exam Tip: How to Score Full Marks

1. Always justify your classification:

   • ❌ "This is ratio data."
   • ✅ "This is ratio data because it has a true zero (e.g., 0 kg of rice) and equal intervals (1 kg increments), allowing arithmetic operations like calculating the mean order quantity."

2. Use real-world examples:

   • For interval data, cite temperature or IQ scores.
   • For discrete data, use counts (e.g., "number of Ncell calls").

3. Visual aids in answers:

   • Draw a table for classification (like the one above).
   • Sketch a bar chart for qualitative data or a histogram for quantitative data.

4. Watch out for trick questions:

   • Age groups (e.g., 18–25, 26–35) are ordinal, not interval, because the intervals (7 years vs. 9 years) are unequal.
   • Years (e.g., 2000 vs. 2023) are interval, not ratio, because there’s no true zero (year 0 doesn’t mean "no time").

5. Practice past exam questions:

   • For Q1, D7, P58 (percentiles/deciles), always convert grouped data to cumulative frequencies first.
   • For moments about an arbitrary point, use the formula: $${\mu }_k^{\prime }=\frac{\sum f(x-A{)}^k}{N}$$ where $A$ is the arbitrary point (e.g., 4 in the past exam).

A table with formulas for mean, median, mode per scale type (e.g., "Mean: Only for interval/ratio").