Measurement Scales & Data Types: Levels, Tools & Applications

TAKEAWAYS:

• Measurement scales (nominal, ordinal, interval, ratio) determine how data is categorized, ordered, or quantified, directly impacting statistical analysis.
• Data types (qualitative vs. quantitative) dictate the research tools (surveys, experiments) and analysis methods (descriptive vs. inferential statistics).
• Nominal scales (e.g., gender, brand names) use labels only, while ratio scales (e.g., income, age) allow true zero and mathematical operations.
• Secondary data (e.g., NEPSE stock prices, Daraz sales reports) must be validated for reliability before use in research.
• Measurement errors (systematic vs. random) threaten validity; tools like Likert scales or pilot tests mitigate these risks.
• Ethical data collection (e.g., anonymizing survey responses) is critical when using scales like ordinal rankings in sensitive topics (e.g., employee satisfaction).

1. Measurement Scales: The Foundation of Data Classification

Measurement scales define how data is categorized, ordered, or quantified. They determine the type of statistical analysis you can perform. The four scales, from lowest to highest precision, are:

1.1 Nominal Scale: Labels Without Order

• Definition: Data is divided into mutually exclusive categories with no inherent order.
• Examples:
  • Gender (Male/Female/Other)
  • Brand preferences (Nepal Telecom vs. Ncell)
  • Political affiliation (CPN-UML, Nepali Congress, Others)
• Operations Allowed:
  • Counting frequencies (e.g., "How many respondents chose Daraz?").
  • No arithmetic operations (e.g., you cannot say "Daraz is twice as good as Pathao").
• Real-World Use:
  • eSewa: Categorizing users by service type (electricity, water, traffic fine).
  • Khalti: Classifying transactions by payment method (credit/debit card, mobile wallet, bank transfer).

1.2 Ordinal Scale: Rankings with Unknown Intervals

• Definition: Data has categories with a meaningful order, but the difference between ranks is not quantifiable.
• Examples:
  • Customer satisfaction surveys (Poor → Fair → Good → Excellent).
  • Employee performance ratings (1 = Poor, 5 = Outstanding).
  • Traffic congestion levels (Low/Medium/High).
• Operations Allowed:
  • Ranking (e.g., "Pathao is ranked higher than Yeti in customer satisfaction").
  • No calculation of differences (e.g., you cannot say "Good is 2x better than Fair").
• Real-World Use:
  • NTC: Ranking cities by internet speed (Kathmandu > Pokhara > Biratnagar).
  • NEPSE: Classifying stocks by market capitalization (Large-Cap, Mid-Cap, Small-Cap).

WORKED EXAMPLE: A researcher asks 100 students to rank their preferred online shopping platforms (Daraz, Sastodeal, Hamrobazaar) from 1 (Least preferred) to 3 (Most preferred). The results:

Analysis:

• Daraz is the most preferred (60 students ranked it #1).
• Cannot conclude: Daraz is "3x better" than Hamrobazaar (ordinal data lacks interval precision).

1.3 Interval Scale: Equal Intervals, No True Zero

• Definition: Data has ordered categories with equal intervals, but no true zero point (zero is arbitrary).
• Examples:
  • Temperature in °C or °F (0°C ≠ "no temperature").
  • IQ scores (100 is average, but 0 does not mean "no intelligence").
  • Likert scale (Strongly Disagree → Strongly Agree, with equal steps).
• Operations Allowed:
  • Addition/subtraction (e.g., "The difference between 20°C and 10°C is 10°C").
  • No multiplication/division (e.g., "20°C is not twice as hot as 10°C").
• Real-World Use:
  • WhatsApp Business: Measuring customer satisfaction on a 1–5 scale.
  • YouTube: Rating videos (1–5 stars) to analyze trends.

WORKED EXAMPLE: A bank (e.g., Nabil Bank) surveys customers on their satisfaction with online banking services using a 5-point Likert scale:

Analysis:

• Mean score for user-friendliness = (5×1 + 10×2 + 30×3 + 40×4 + 15×5) / 100 = 3.5 (slightly positive).
• Cannot say: "User-friendliness is 3.5x better than neutral" (interval data lacks a true zero).

1.4 Ratio Scale: True Zero and Full Mathematical Operations

• Definition: Data has ordered categories with equal intervals and a true zero (zero means "nothing").
• Examples:
  • Age, income, weight, number of employees.
  • Sales revenue (₹0 = no sales).
  • Time taken to complete a task.
• Operations Allowed:
  • All mathematical operations (addition, subtraction, multiplication, division).
  • Calculating ratios (e.g., "Company A’s revenue is 2x that of Company B").
• Real-World Use:
  • Daraz: Analyzing sales growth (₹500M in 2022 vs. ₹1B in 2023).
  • NEPSE: Comparing market capitalization of companies (e.g., NMB vs. Global IME).

WORKED EXAMPLE: A Chaudhary Group subsidiary tracks employee productivity (hours worked per week):

Analysis:

• Logistics employees work 1.37x more hours than Retail (52/38 ≈ 1.37).
• Can conclude: Productivity differences are meaningful (true zero exists).

2. Data Types: Qualitative vs. Quantitative

Data types determine how you collect and analyze information. The two broad categories are:

mindmap
  root((Data Types))
    Qualitative
      [[Customer Reviews]]
        - Daraz: "Fast delivery, but packaging weak"
        - Pathao: "Riders are polite"
      [[Focus Groups]]
        - Himalayan Java: Barista training feedback
    Quantitative
      [[Surveys]]
        - NTC: "How many hours do you use mobile data daily?"
          - Nominal: [GPRS, 3G, 4G, 5G]
          - Ratio: [1-10 hours]
      [[Experiments]]
        - Nabil Bank: A/B testing of app UI
          - Conversion rates: 5% (Old UI) vs. 12% (New UI)
      [[Secondary Data]]
        - NEPSE: Stock prices (₹1000 → ₹1500 in 6 months)

2.1 Qualitative Data: The Voice of Respondents

• Sources:
  • Open-ended survey questions (e.g., "Why did you choose Khalti over eSewa?").
  • Interviews, focus groups, case studies.
• Advantages:
  • Reveals why behind behaviors (e.g., "Customers prefer Daraz for its wide product range").
  • Useful for exploratory research (e.g., "What challenges do SMEs face in Nepal?").
• Disadvantages:
  • Subjective (biased by researcher interpretation).
  • Time-consuming to analyze.
• Real-World Use:
  • Pathao: Conducting interviews with drivers to improve app usability.
  • Himalayan Java: Analyzing customer feedback on coffee flavors (e.g., "Too bitter").

WORKED EXAMPLE: A researcher studies why students prefer online vs. offline shopping in Kathmandu. Qualitative data collected:

• "Online shopping saves time; I don’t have to stand in traffic."
• "I trust offline stores because I can see the product before buying."

Analysis:

• Themes identified:
  1. Convenience (online).
  2. Trust (offline).
• Cannot quantify: Cannot say "70% prefer online for convenience" (requires surveys).

2.2 Quantitative Data: Numbers That Speak

• Sources:
  • Closed-ended surveys (Likert scales, multiple-choice).
  • Experiments, secondary data (e.g., NEPSE stock prices, NTC internet speed reports).
• Advantages:
  • Objective and easy to analyze (statistical software like SPSS).
  • Generalizable (e.g., "60% of Nepali users prefer mobile wallets").
• Disadvantages:
  • Lacks context (e.g., "Sales increased by 20%" but why?).
  • May miss nuances (e.g., a "5" on a Likert scale could mean "satisfied" or "neutral").
• Real-World Use:
  • Khalti: Tracking transaction volumes (₹500M/day in 2023).
  • Ncell: Measuring customer churn rate (5% monthly).

WORKED EXAMPLE: A Nabil Bank study collects quantitative data on loan defaults:

Analysis:

• Highest default rate: Business loans (12%).
• Statistical test: Chi-square to check if loan type affects defaults.
• Limitation: Does not explain why business loans default more (requires qualitative data).

3. Secondary Data: The Hidden Goldmine

Secondary data is previously collected data used for new research. Sources include:

• Government: NEPSE reports, CBS Nepal census data.
• Private: Daraz sales reports, Khalti transaction logs.
• Academic: Journal articles, past theses.

Advantages of Secondary Data

• Cost-effective (no need for new surveys).
• Time-saving (data already exists).
• Broader scope (e.g., NTC’s nationwide internet speed data).

Disadvantages of Secondary Data

WORKED EXAMPLE: A student researches "Factors affecting e-commerce growth in Nepal" and uses:

1. Primary data: Surveys from 200 Daraz customers.
2. Secondary data:
   • NEPSE’s 2023 e-commerce revenue report (₹200B).
   • CBS Nepal’s 2022 internet penetration data (30% urban, 10% rural).

Validation Check:

• Reliability: Cross-check NEPSE data with IMF reports for consistency.
• Relevance: Ensure the data is from 2022–2023 (not older than 2 years).

4. Measurement Errors: The Silent Threat

Errors distort data and lead to invalid conclusions. Two types:

Real-World Case:

• eSewa’s 2021 outage: Due to server overload, transaction data was underreported by 15% (systematic error).
• Fix: eSewa now uses redundant servers to prevent data loss.

5. Ethical Considerations in Measurement

Ethics ensure respect for participants and data integrity. Key concerns:

1. Anonymity/Confidentiality:
   • Example: A Ncell survey on customer satisfaction must not link responses to individual SIM numbers.
2. Informed Consent:
   • Example: Before recording focus group discussions on Daraz’s return policy, participants must agree in writing.
3. Avoiding Harm:
   • Example: Not asking sensitive questions (e.g., income) unless necessary for the study.

Case Study: Unethical Scenario A group of students plans to research "fetal abnormalities via ultrasound" without:

• Parental consent.
• Explaining risks to pregnant women.
• Ensuring data privacy.

Why it’s unethical:

• Violates autonomy (participants not fully informed).
• Risks psychological harm (stress from test results).
• Data misuse (ultrasound images could be shared without consent).

In the Real World

1. Khalti’s Transaction Data (Ratio Scale)

   • How it’s used: Khalti tracks transaction amounts (₹) to analyze spending patterns (e.g., "₹500M spent on groceries in Jan 2024").
   • Scale applied: Ratio (₹0 = no transaction; can calculate growth rates).
   • Business impact: Identifies peak spending times (e.g., Dashain) to offer discounts.

2. Daraz’s Customer Satisfaction Survey (Ordinal + Interval)

   • How it’s used: Daraz uses a 5-point Likert scale (1 = Poor, 5 = Excellent) to measure:
     • Delivery speed.
     • Product quality.
   • Scales applied:
     • Ordinal: Rankings (e.g., "Top 3 complaints").
     • Interval: Mean scores (e.g., "Delivery speed = 3.8/5").
   • Real impact: Low scores on return policy led to faster refunds.

3. NEPSE’s Stock Market Data (Ratio Scale)

   • How it’s used: NEPSE provides daily stock prices (₹) to analyze:
     • Company performance (e.g., NMB vs. Global IME).
     • Market trends (e.g., "Small-cap stocks grew 15% in 2023").
   • Scale applied: Ratio (₹0 = no value; can calculate P/E ratios).
   • Investor use: Helps traders decide buy/sell based on historical data.

Exam Tip

1. Scale Hierarchy: Remember the order of scales (Nominal → Ordinal → Interval → Ratio) and what operations each allows.

   • Common mistake: Saying "ordinal scales allow multiplication" (they don’t!).

2. Data Type Matching:

   • Qualitative → Themes, quotes.
   • Quantitative → Statistics (mean, standard deviation).
   • Exam trick: If a question asks for "statistical analysis", the data must be quantitative (interval/ratio).

3. Secondary Data Questions:

   • Always validate secondary data by checking:
     • Source credibility (e.g., CBS Nepal > random blog).
     • Recency (data from 2022 is better than 2018 for 2024 exams).
     • Relevance (e.g., NTC’s internet speed data is useless for a customer loyalty study).

4. Case Study Approach:

   • For application-based questions (e.g., "How would you measure customer satisfaction for Pathao?"):
     • Step 1: Identify the scale (Likert = interval).
     • Step 2: Choose the data type (quantitative for scores).
     • Step 3: Justify why (e.g., "Interval allows us to calculate average satisfaction").

5. Ethics Shortcuts:

   • If a scenario involves sensitive data (health, finance), always mention:
     • Anonymity.
     • Consent.
     • Right to withdraw.

mindmap
  root((Measurement Scales & Data Types))
    Nominal
      "Labels only (Gender, Brands)"
      "No order, no math"
    Ordinal
      "Rankings (1st, 2nd, 3rd)"
      "No interval precision"
    Interval
      "Equal intervals (Temperature, Likert)"
      "No true zero"
    Ratio
      "True zero (Income, Age)"
      "Full math operations"
    Qualitative
      "Descriptive (Interviews, Themes)"
      "Nominal/Ordinal scales"
    Quantitative
      "Numeric (Surveys, Experiments)"
      "Interval/Ratio scales"
    Secondary Data
      "Pre-collected (NEPSE, CBS)"
      "Validate before use!"
    Errors
      "Systematic vs. Random"
      "Mitigate with calibration/pilot tests"