Measurement Scales & Tools: Types, Validity, Reliability

TAKEAWAYS:

• Measurement scales classify data from categorical (nominal) to quantitative (ratio), each with distinct uses and math rules.
• Validity ensures a tool measures what it claims (e.g., a happiness survey’s questions), while reliability ensures consistent results (e.g., NTC’s signal strength tests).
• Nominal scales (e.g., gender, brand names) label only; ordinal scales (e.g., customer satisfaction: poor/good/excellent) rank but don’t quantify gaps.
• Interval and ratio scales enable statistical analysis: interval (e.g., temperature in °C) lacks a true zero, while ratio (e.g., income, stock prices) does.
• Tools like Likert scales (1–5 ratings) or structured questionnaires rely on scale choice to avoid bias (e.g., Daraz’s order priority surveys).
• Real-world tie: NEPSE’s stock indices use ratio scales to calculate percentage gains/losses, while Pathao’s ride ratings use ordinal scales for driver performance.

1. Definitions: What Is Measurement in Research?

Measurement is the process of assigning numbers or labels to observations to describe variables objectively. In business research, it bridges raw data (e.g., customer feedback) with meaningful insights (e.g., market trends). Without proper scales, data becomes ambiguous—e.g., calling "high" and "low" satisfaction scores without defining their intervals leads to misinterpretation.

Key Terms:

• Variable: Any trait or characteristic that can take different values (e.g., age, sales revenue, customer loyalty).
• Scale: A rule or standard that assigns numbers/labels to variables (e.g., Likert scale for agreement levels).
• Measurement Error: Discrepancy between true value and observed value (e.g., a thermometer reading 25°C when it’s actually 24°C).

2. Types of Measurement Scales: A Hierarchical Classification

Scales are categorized by their level of measurement, which determines the statistical operations they allow. Below is a mindmap of the four primary scales, ordered from lowest to highest precision:

mindmap
  root((Measurement Scales))
    Nominal["Labels only (no order)
    - Example: Gender (Male/Female), Brand (Ncell/eSewa)
    - Stats: Mode only"]
    Ordinal["Ranked categories (no equal intervals)
    - Example: Survey ratings (Poor/Good/Excellent), Pathao ride ratings (1-5 stars)
    - Stats: Mode, Median"]
    Interval["Equal intervals, no true zero
    - Example: Temperature (°C), IQ scores
    - Stats: Mean, Standard Deviation"]
    Ratio["Equal intervals + true zero
    - Example: Income (₹10,000), Stock prices (NEPSE index), Age (25 years)
    - Stats: All (mean, ratio, regression)]

Visual Comparison Table:

3. Nominal Scale: Categorizing Without Order

Definition: Assigns unique labels to categories with no numerical meaning or order. Think of it as a "name tag" for data.

How It Works:

• Categories are mutually exclusive (no overlap) and exhaustive (cover all possibilities).
• Example: Survey question: "What payment method do you use?"
  • Options: Debit Card (Ncell eWallet), Credit Card, Bank Transfer, Cash
  • Not: "High/Medium/Low" (this would be ordinal).

Real-World Example: eSewa vs. Khalti Market Share

• Measurement: Users are categorized into eSewa users or Khalti users (nominal scale).
• Analysis: Researchers count frequencies to determine which platform is more popular in Kathmandu.
• Limitation: Cannot compare "how much more" one is used than the other (e.g., cannot say Khalti is "twice as popular").

Worked Example: A study asks 200 Nepali shoppers: "Which online grocery platform do you use?"

• Data: 120 (Daraz), 50 (Sasto), 30 (NepalGrocery)
• Analysis: Daraz has the highest mode (most frequent category), but we cannot calculate averages or ratios.

4. Ordinal Scale: Ranking Without Intervals

Definition: Categories have a natural order, but the distance between ranks is unknown. Example: Likert scale ratings.

How It Works:

• Respondents rank items from lowest to highest (e.g., "Strongly Disagree" to "Strongly Agree").
• Cannot assume equal intervals: A "4" on a 5-point scale is not necessarily twice a "2."

Example: Pathao Driver Ratings

• Survey question: "How would you rate your last Pathao ride?"
  • Options: 1 (Poor) → 2 (Fair) → 3 (Good) → 4 (Very Good) → 5 (Excellent)
• Analysis: Drivers with a median rating of 4 are performing better than those with a median of 2, but we cannot say the difference between 4 and 5 is the same as between 2 and 3.

Mermaid Diagram: Likert Scale Structure

flowchart TD
    A["Strongly Disagree"] --> B["Disagree"]
    B --> C["Neutral"]
    C --> D["Agree"]
    D --> E["Strongly Agree"]

Advantages:

• Captures qualitative order (e.g., customer preference trends).
• Useful for ranking (e.g., "Top 3 reasons for using eSewa").

Disadvantages:

• No arithmetic operations: Cannot calculate means or standard deviations.
• Subjective interpretation: A "4" might mean different things to different people.

5. Interval Scale: Equal Intervals, No True Zero

Definition: Categories have equal intervals between values, but no absolute zero (e.g., 0°C does not mean "no temperature").

How It Works:

• Differences between values are meaningful (e.g., 20°C is 10°C warmer than 10°C).
• No ratios: You cannot say 40°C is "twice as hot" as 20°C (it’s actually hotter by 20°C).

Example: Temperature Data for Kathmandu’s Tourism

• Measurement: Monthly average temperatures (°C).
  • January: 12°C, February: 15°C, March: 20°C.
• Analysis:
  • The mean temperature in March is higher than January by 8°C.
  • Standard deviation can measure variability, but ratios are invalid (e.g., "March is not twice as warm as January").

Real-World Tie: NEPSE Stock Index

• Measurement: Stock prices are often measured on an interval scale (e.g., NEPSE index points).
• Example: If NEPSE rises from 1,200 to 1,400 points, the increase is 200 points, but we cannot say it’s "twice as valuable" as 1,200.

6. Ratio Scale: The Gold Standard for Quantitative Data

Definition: Combines equal intervals with a true zero, allowing the most statistical operations (ratios, percentages, etc.).

How It Works:

• Zero means absence of the variable (e.g., 0 income = no income).
• Ratios are meaningful: A salary of ₹50,000 is twice ₹25,000.

Examples:

• Age: 30 years is double 15 years.
• Income: A ₹10,000 loan is 10% of a ₹100,000 salary.
• Stock Prices: If NEPSE rises from 1,000 to 1,200, it’s a 20% increase.

Worked Example: Loan Interest Calculation A bank offers a loan with simple interest:

• Principal (P) = ₹50,000
• Rate (r) = 10% per annum
• Time (t) = 2 years
• Interest (I) = P × r × t = ₹50,000 × 0.10 × 2 = ₹10,000
• Total repayment = ₹60,000
• Analysis: The ratio of interest to principal is 20% (₹10,000/₹50,000), which is only possible with a ratio scale.

7. Validity vs. Reliability: The Twin Pillars of Measurement

Both concepts ensure trustworthy data, but they measure different aspects:

flowchart TD
    A["Measurement Quality"] --> B["Validity['Measures what it claims to measure']"]
    A --> C["Reliability['Consistently measures the same thing']"]
    B --> D["Example: A happiness survey’s questions accurately capture well-being."]
    C --> E["Example: NTC’s signal strength test gives the same result twice."]

Table: Key Differences

Real-World Example: NEPSE’s Stock Index Validity

• Validity Check: Does NEPSE’s index truly reflect market performance?
  • Yes, if it includes a representative sample of listed companies (e.g., Nabil Bank, Himalayan Java).
  • No, if it excludes small-cap stocks, skewing results toward large corporations.
• Reliability Check: Does NEPSE’s index stay consistent over time?
  • Yes, if daily calculations use the same methodology (e.g., market capitalization weighting).

8. Measurement Tools: Choosing the Right Instrument

The scale type dictates the tool used to collect data. Below is a decision flowchart for selecting tools:

flowchart TD
    A["What scale do I need?"] --> B{"Is it categorical?"}
    B -->|"Yes"| C["Use: Nominal/Ordinal Scale Tools"]
    B -->|"No"| D["Use: Interval/Ratio Scale Tools"]
    C --> E["Questionnaires with multiple-choice/likert items"]
    C --> F["Observational checklists (e.g., counting Ncell vs. NTC users)"]
    D --> G["Surveys with numerical ratings"]
    D --> H["Experiments with ratio data (e.g., sales revenue)"]

Common Tools by Scale Type:

Worked Example: Designing a Questionnaire for Khalti’s User Satisfaction Objective: Measure user satisfaction with Khalti’s transaction speed (ordinal scale). Tool: 5-point Likert scale:

1. Very Slow
2. Slow
3. Neutral
4. Fast
5. Very Fast Question: "How would you rate Khalti’s transaction processing time?"

• Scale Type: Ordinal (ranks but no equal intervals).
• Analysis: Calculate the median satisfaction score to identify trends (e.g., 70% of users rate it "Fast" or better).

9. Challenges in Measurement: Pitfalls to Avoid

Even with the right scales, researchers face common issues:

1. Misclassification:

   • Problem: Assigning an interval scale to ordinal data (e.g., treating Likert ratings as numerical for averages).
   • Fix: Use non-parametric tests (e.g., median instead of mean) for ordinal data.

2. Response Bias:

   • Problem: Survey respondents may lie or misremember (e.g., overestimating eSewa usage).
   • Fix: Use anonymous surveys or observational data (e.g., transaction logs).

3. Scale Ambiguity:

   • Problem: Vague labels (e.g., "Good" vs. "Very Good") lead to inconsistent interpretations.
   • Fix: Define anchors clearly (e.g., "Good = 3-4 transactions per day").

4. Tool Limitations:

   • Problem: A questionnaire may not capture why users prefer Khalti over eSewa (only what they prefer).
   • Fix: Combine quantitative (scales) with qualitative (open-ended questions) methods.

10. In the Real World

Measurement scales and tools are everywhere in Nepal’s business ecosystem. Here’s how companies apply them:

1. Daraz: Order Priority Queues (Ordinal Scale)

• Idea: Daraz uses ordinal scales to rank customer orders by urgency (e.g., "Standard," "Express," "Same-Day").
• How:
  • Customers select a delivery speed (1 = Slowest, 3 = Fastest).
  • Daraz’s logistics team prioritizes orders based on this ranking.
• Worked Example:
  • Data: 60% of orders are "Standard" (rank 2), 30% "Express" (rank 3), 10% "Same-Day" (rank 1).
  • Analysis: The median rank is 2, showing most users prefer balanced speed vs. cost.

2. NEPSE: Stock Market Indices (Ratio Scale)

• Idea: NEPSE’s Composite Index uses a ratio scale to measure stock performance.
• How:
  • Index value = (Current market cap / Base market cap) × Base index (e.g., 1,000).
  • A rise from 1,200 to 1,400 points is a 16.67% increase (ratio calculation).
• Real Impact:
  • Investors use this to decide whether to buy/sell stocks (e.g., Nabil Bank shares).
  • Policymakers monitor trends to adjust economic regulations.

3. NTC/Ncell: Signal Strength Surveys (Interval Scale)

• Idea: Mobile operators use interval scales to measure signal coverage (e.g., -50 dBm to -100 dBm).
• How:
  • Surveys ask users to rate signal strength on a 5-point scale (Excellent to Poor).
  • Conversion: Each point is assigned an interval value (e.g., 1 = -100 dBm, 5 = -50 dBm).
• Worked Example:
  • Data: 40% of users rate signal as "Good" (3/5), 30% "Fair" (2/5).
  • Analysis: The mean interval value is 2.8, indicating moderate coverage in Kathmandu’s Valley.

11. Exam Tip: How to Score Full Marks

This unit tests conceptual understanding and application. Follow these strategies:

1. Differentiate Validity vs. Reliability (20 Marks)

• Validity: "Does the tool measure what it’s supposed to?"
  • Example: A survey on "customer loyalty" must ask about repeat purchases, not just satisfaction.
• Reliability: "Does the tool give the same result repeatedly?"
  • Example: A stopwatch should record the same time for a 10-second sprint in two trials.
• Common Mistake: Confusing the two (e.g., saying a thermometer is reliable if it’s always broken).
• Fix: Use real-world analogies:
  • "A doctor’s blood pressure cuff is reliable if it gives the same reading twice, but valid only if it measures blood pressure (not pulse)."

2. Explain Scale Types with Examples (30 Marks)

• Structure your answer like this:
  1. Define the scale (e.g., "Nominal scales assign labels without order").
  2. Give 1 example (e.g., "Gender: Male/Female").
  3. State allowed stats (e.g., "Mode only").
  4. Provide a Nepali business use (e.g., "eSewa vs. Khalti user segmentation").
• Avoid: Listing scales without examples or using the wrong stats (e.g., saying "mean" is allowed for ordinal data).

3. Design a Questionnaire for a Given Scale (20 Marks)

• Steps:
  1. Identify the scale needed (e.g., ordinal for satisfaction).
  2. Choose the right tool (e.g., Likert scale for 1-5 ratings).
  3. Write 2-3 questions with clear anchors (e.g., "How would you rate Pathao’s driver courtesy?").
  4. Justify your choice (e.g., "Ordinal scale is used because we rank drivers but cannot quantify ‘how much’ courtesy").
• Example Answer:

  *"To measure customer satisfaction with Nabil Bank’s ATM services (ordinal scale), I would use a 5-point Likert scale:

  1. Very Dissatisfied
  2. Dissatisfied
  3. Neutral
  4. Satisfied
  5. Very Satisfied Justification: This scale ranks satisfaction levels but does not assume equal intervals between points."*

4. Solve a Worked Example (10 Marks)

• Approach:
  1. Identify the scale in the data (e.g., "Income data is ratio scale").
  2. Calculate the right statistic (e.g., mean income for ratio data).
  3. Interpret results in context (e.g., "The average income of ₹30,000 suggests middle-class customers for Himalayan Java’s products").
• Example:

  "Given the following customer ages for a Daraz survey: 25, 30, 35, 40, 45 (ratio scale), the mean age is 35. This indicates Daraz’s primary user is in the 30-40 age group, ideal for targeted promotions."

5. Avoid Common Pitfalls

• Do not:
  • Mix up scales (e.g., treating nominal data as interval).
  • Use ratio stats for interval data (e.g., calculating ratios with temperature).
  • Ignore validity/reliability in tool selection.
• Do:
  • Always justify your scale choice.
  • Use Nepali examples (e.g., NEPSE, Daraz, eSewa) to show real-world application.
  • Draw flowcharts or tables to compare scales (exams love visuals!).

12. Summary Mindmap: Key Takeaways

mindmap
  root((Unit 9: Measurement Scales & Tools))
    Scales["Four Types"]
      Nominal["Labels only (eSewa/Khalti users)"]
      Ordinal["Ranked (Pathao ratings 1-5)"]
      Interval["Equal intervals (NEPSE index points)"]
      Ratio["True zero (income, age)"]
    Validity["Measures what it claims"]
    Reliability["Consistent results"]
    Tools["Questionnaires, surveys, experiments"]
    Challenges["Misclassification, bias, ambiguity"]
    RealWorld["NEPSE (ratio), Daraz (ordinal), NTC (interval)"]