Sampling Design: Types, Methods & Applications

TAKEAWAYS:

• Sampling design ensures research findings are representative, cost-effective, and generalizable by selecting a subset of the population.
• Probability sampling (random, stratified, cluster) guarantees statistical validity, while non-probability sampling (convenience, quota, snowball) is faster but biased.
• Sample size depends on population size, confidence level, and margin of error—use Krejcie & Morgan’s table or Sloven’s formula for calculations.
• Ethical sampling requires informed consent, anonymity, and avoidance of coercion (e.g., NTC’s customer satisfaction surveys).
• Real-world applications: Nepal Rastra Bank uses stratified sampling for financial inclusion studies, while Daraz employs convenience sampling for quick feedback.
• Common mistakes: Under-sampling, over-representing a group, or ignoring non-response bias—all lead to invalid conclusions.

1. What is Sampling Design?

Sampling design is the process of selecting a subset (sample) from a larger group (population) to study, ensuring the sample accurately represents the population’s characteristics. Poor sampling leads to biased or unreliable research.

Key Terms:

2. Why is Sampling Important?

• Cost & Time Efficiency: Studying 1,000 out of 10 million Daraz users is feasible.
• Practicality: Some populations (e.g., all Nepali voters) are too large to survey fully.
• Accuracy: Proper sampling reduces margin of error (e.g., ±3% in election polls).
• Generalizability: Findings from a well-chosen sample apply to the entire population.

3. Probability vs. Non-Probability Sampling

4. Probability Sampling Methods

(A) Simple Random Sampling

• Every member has an equal chance of selection (e.g., lottery system).
• How it works:
  1. Define population (e.g., all Nabil Bank customers).
  2. Assign numbers (e.g., 1 to 10,000).
  3. Use random number generator to pick 500 customers.

Advantages: ✅ Unbiased (no researcher influence). ✅ Easy to analyze (statistical tests like t-tests work).

Disadvantages: ❌ Time-consuming (e.g., contacting 500 random Ncell users). ❌ Hard to get a complete sampling frame (e.g., unlisted landline numbers).

Real-World Example:

• Nepal Electricity Authority (NEA) uses simple random sampling to select households for electricity bill audits.

(B) Stratified Sampling

• Population is divided into subgroups (strata) based on shared traits, then randomly sampled from each.
• Example: Studying customer satisfaction at Nabil Bank → Strata:
  • Salaried employees (30%)
  • Business owners (40%)
  • Students (20%)
  • Retirees (10%)

Steps:

1. Divide population into homogeneous groups.
2. Randomly sample from each stratum proportionally.

Advantages: ✅ More precise than simple random sampling. ✅ Ensures representation of all groups (e.g., gender, income levels).

Disadvantages: ❌ Complex (requires accurate strata definitions). ❌ Costly (e.g., separate surveys for each stratum).

Real-World Example:

• Nepal Rastra Bank uses stratified sampling to study financial literacy across urban, rural, and remote areas.

(C) Cluster Sampling

• Population is divided into clusters (e.g., schools, districts), then entire clusters are randomly selected.
• Example: Studying traffic congestion in Kathmandu → Clusters:
  • Lalitpur (Cluster 1)
  • Bhaktapur (Cluster 2)
  • Kageshwori (Cluster 3)

Steps:

1. Identify clusters (e.g., wards in Kathmandu).
2. Randomly select clusters (e.g., 5 out of 10 wards).
3. Survey all members in selected clusters.

Advantages: ✅ Cost-effective (e.g., surveying all households in one cluster). ✅ Useful for large/geographically spread populations (e.g., Nepal’s rural areas).

Disadvantages: ❌ Less precise than stratified sampling. ❌ Clusters may not be homogeneous (e.g., one ward is wealthy, another is poor).

Real-World Example:

• Central Bureau of Statistics (CBS) uses cluster sampling for household income surveys in Nepal.

5. Non-Probability Sampling Methods

(A) Convenience Sampling

• Selects easiest-to-reach members (e.g., students in a class, shopping mall visitors).
• Example: Daraz asks first 100 customers who visit their booth at Thapathali for feedback.

Advantages: ✅ Fast & cheap. ✅ No sampling frame needed.

Disadvantages: ❌ Highly biased (e.g., only tech-savvy users respond). ❌ Not generalizable.

Real-World Example:

• Pathao uses convenience sampling to get driver feedback at pickup points.

(B) Quota Sampling

• Predefined quotas for subgroups (e.g., 30% male, 70% female).
• Example: Nepal Telecom (NTC) wants 500 responses → 250 males, 250 females.

Steps:

1. Define quotas (e.g., age, gender, location).
2. Stop sampling once quotas are filled (even if non-random).

Advantages: ✅ Ensures representation of key groups. ✅ Faster than probability methods.

Disadvantages: ❌ Researcher bias (e.g., only interviewing willing participants). ❌ Not statistically reliable.

Real-World Example:

• Khalti uses quota sampling to balance feedback from urban vs. rural users.

(C) Snowball Sampling

• Initial respondents refer others (useful for hard-to-reach groups).
• Example: Studying undocumented migrant workers in Kathmandu.

Steps:

1. Find 1-2 initial respondents.
2. Ask them to refer others with similar traits.

Advantages: ✅ Useful for hidden/private populations. ✅ Low cost.

Disadvantages: ❌ Highly biased (e.g., only connected individuals respond). ❌ Non-generalizable.

Real-World Example:

• NGOs studying child labor use snowball sampling to find former child workers.

6. Determining Sample Size

Sample size affects accuracy and cost. Use these formulas:

(A) Slovin’s Formula (for finite populations)

$$n=\frac{N}{1+N(e{)}^2}$$

• $n$ = Sample size
• $N$ = Population size
• $e$ = Margin of error (e.g., 0.05 for 5%)

Example:

• Population (N): 10,000 Nepal Rastra Bank customers.
• Margin of error (e): 5% (0.05).
• Calculation: $$n=\frac{10,000}{1+10,000\times (0.05{)}^2}=\frac{10,000}{1+25}=\frac{10,000}{26}\approx 385$$
• Sample size needed: 385 customers.

(B) Krejcie & Morgan’s Table (Quick Lookup)

Example:

• Population: 5,000 Daraz sellers.
• Sample size: 370 sellers.

7. Ethical Considerations in Sampling

• Informed Consent: Participants must know they’re being studied (e.g., Nepal Rastra Bank’s survey disclosures).
• Anonymity/Confidentiality: No personal data should be traceable (e.g., Khalti’s feedback forms).
• Avoid Coercion: No pressure to participate (e.g., NTC not forcing customers to take surveys).
• Fair Representation: No under/over-sampling of groups (e.g., not ignoring rural areas in a national study).

Real-World Case Study: Nepal Rastra Bank’s Financial Inclusion Survey

• Method: Stratified random sampling (urban, rural, remote).
• Ethics:
  • Informed consent via written disclaimers.
  • Anonymity ensured by coding responses.
  • No coercion—participants voluntarily joined.

8. Common Sampling Mistakes & How to Avoid Them

In the Real World

1. Nepal Rastra Bank (NRB) Financial Inclusion Surveys

   • Method: Stratified sampling (urban, rural, remote districts).
   • Why? Ensures all economic groups (rich, poor, middle-class) are represented.
   • Impact: Helps design microfinance policies for unbanked populations.

2. Daraz Customer Feedback System

   • Method: Convenience sampling (pop-up surveys at checkout).
   • Why? Fast and cheap, but biased toward tech-savvy users.
   • Improvement: Could use stratified sampling (e.g., rural vs. urban buyers).

3. Nepal Electricity Authority (NEA) Bill Audits

   • Method: Cluster sampling (selecting entire wards randomly).
   • Why? Cost-effective for large, spread-out populations.
   • Challenge: Wards vary in income—may need stratification for accuracy.

4. Pathao Driver Satisfaction Study

   • Method: Snowball sampling (happy drivers refer others).
   • Why? Hard to reach all drivers directly.
   • Risk: Only satisfied drivers respond—misses complaints.

5. Nepal Stock Exchange (NEPSE) Investor Sentiment Polls

   • Method: Quota sampling (equal shares for institutional, retail, foreign investors).
   • Why? Ensures no group dominates the survey.
   • Ethical Issue: Incentivizing responses (e.g., discounts on trades) can bias results.

Exam Tip

How This Unit is Tested in PU Exams

1. Definitions & Differences (5-10 marks):

   • Expect short-answer questions like:
     • "Differentiate between stratified and cluster sampling with examples."
     • "What is sampling error? How does it differ from non-sampling error?"

2. Scenario-Based Questions (10-15 marks):

   • Case studies where you must choose the best sampling method.
   • Example:

     "A researcher wants to study traffic congestion in Kathmandu. The population is 2 million vehicles. Suggest a sampling method and justify your choice."

3. Calculations (5-10 marks):

   • Sample size determination using Slovin’s formula or Krejcie & Morgan’s table.
   • Example:

     "Calculate the sample size for a population of 5,000 with a 5% margin of error."

4. Ethical & Practical Issues (5 marks):

   • Short notes on:
     • "What are the ethical concerns in snowball sampling?"
     • "How can non-response bias be minimized in a mail survey?"

5. Real-World Application (5-10 marks):

   • Apply sampling methods to Nepali businesses (e.g., Nepal Rastra Bank, Daraz, NTC).
   • Example:

     "How would you design a sample to study customer satisfaction at Nabil Bank? Justify your method."

How to Score Full Marks

✔ Use real-world examples (e.g., Nepal Rastra Bank, Daraz, NTC). ✔ Compare methods in tables (e.g., probability vs. non-probability). ✔ Show calculations (e.g., sample size formula). ✔ Discuss advantages/disadvantages with Nepali context. ✔ Link to ethics (e.g., informed consent, anonymity).

Avoid: ❌ Vague answers (e.g., "Sampling is important"—explain why). ❌ Ignoring biases (e.g., "Convenience sampling is fine"—mention limitations). ❌ Wrong formulas (e.g., using mean formula instead of Slovin’s).

Final Mermaid Summary:

mindmap
  root((Sampling Design))
    Probability Sampling
      Simple Random
      Stratified
      Cluster
    Non-Probability Sampling
      Convenience
      Quota
      Snowball
    Sample Size Determination
      Slovin's Formula
      Krejcie & Morgan Table
    Ethical Considerations
      Informed Consent
      Anonymity
      Avoid Coercion
    Real-World Applications
      Nepal Rastra Bank (Stratified)
      Daraz (Convenience)
      NTC (Quota)