Research Design: Types, Models & Selection Criteria

TAKEAWAYS:

• Research design is the blueprint of a study, determining how data will be collected, analyzed, and interpreted to answer research questions.
• The five main types of research design (exploratory, descriptive, explanatory, experimental, and quasi-experimental) differ in purpose, structure, and control over variables.
• Quantitative, qualitative, and mixed-methods designs serve distinct research goals, with quantitative focusing on measurable data, qualitative on in-depth insights, and mixed-methods combining both.
• Experimental designs (true experiments and quasi-experiments) are used when causality must be established, while non-experimental designs (descriptive, exploratory) are suited for observational studies.
• The selection of a research design depends on research objectives, feasibility, ethical considerations, and the nature of the research problem.
• Real-world applications include A/B testing in apps (e.g., Pathao’s ride pricing models), customer satisfaction surveys (e.g., Daraz’s feedback systems), and clinical trials (e.g., COVID-19 vaccine research).

1. Definition and Purpose of Research Design

Research design is a structured plan that outlines the methods and procedures for collecting, analyzing, and interpreting data to address a research problem. It ensures:

• Validity: The study measures what it claims to measure.
• Reliability: Results can be replicated under similar conditions.
• Objectivity: Minimizes bias and subjectivity.

A well-designed study provides logical flow from research questions to conclusions, ensuring the study is feasible, ethical, and rigorous.

2. Components of Research Design

Every research design includes the following key elements:

a) Research Questions

Research questions are specific, answerable queries that guide the study. They should be:

• Clear and focused (e.g., "How does mobile banking adoption affect financial inclusion in Nepal?").
• Feasible (can be answered with available resources).
• Relevant (addresses a gap in existing knowledge).

Example: If a bank wants to study the impact of interest rate changes on loan repayment, the research question could be: "Does a 2% reduction in interest rates increase loan repayment rates by 15% in urban Nepal?"

b) Hypotheses

A hypothesis is a testable prediction about the relationship between variables. It can be:

• Null hypothesis (H₀): No effect exists (e.g., "Interest rate changes have no effect on loan repayment.").
• Alternative hypothesis (H₁): An effect exists (e.g., "A 2% interest rate reduction increases loan repayment by 15%.").

c) Variables

Variables are factors that can change in a study. They are classified as:

d) Methodology

This includes:

• Data collection methods (surveys, experiments, interviews).
• Sampling techniques (random, stratified, convenience).
• Data analysis methods (statistical tests, thematic analysis).

3. Types of Research Design

Research designs are categorized based on purpose, structure, and control over variables.

a) Exploratory Design

Purpose: To explore a problem, generate ideas, or identify patterns. Characteristics:

• Used when little is known about the topic.
• No hypotheses are tested.
• Qualitative methods (interviews, case studies) are common.

Example: A startup like Pathao might use exploratory design to understand why some drivers accept fewer rides during peak hours. They could conduct interviews with drivers to identify key pain points.

b) Descriptive Design

Purpose: To describe characteristics of a population or situation. Characteristics:

• Answers "what," "how," or "when" questions.
• No manipulation of variables (observational).
• Uses surveys, case studies, or archival data.

Example: Nepal Electricity Authority (NEA) might use descriptive design to analyze electricity consumption patterns across different regions in Nepal. They could collect data on hourly usage, peak demand times, and regional differences.

c) Explanatory Design

Purpose: To explain why or how a phenomenon occurs. Characteristics:

• Answers "why" questions.
• Uses correlational or causal analysis.
• Often involves statistical modeling.

Example: A Daraz study might explore why customer return rates are higher in certain product categories. They could analyze purchase data, product descriptions, and customer reviews to identify trends.

d) Experimental Design

Purpose: To establish causality between variables. Characteristics:

• Manipulation of independent variables.
• Random assignment of subjects to groups.
• Control group (no treatment) vs. experimental group (treatment).

Types of Experimental Designs:

Example (True Experiment): A Nepalese bank tests whether SMS reminders increase loan repayment rates. They randomly assign 1000 borrowers into two groups:

• Group A: Receives SMS reminders.
• Group B: No reminders. After 6 months, they compare repayment rates.

e) Non-Experimental Design

Used when manipulation is not possible (e.g., studying historical events, public opinion). Types:

• Correlational Design: Measures relationships (e.g., "Does higher education level correlate with higher income?").
• Case Study: In-depth analysis of a single case (e.g., "How did Ncell recover from the 2015 earthquake?").
• Survey Design: Collects data from a sample (e.g., "What are the biggest challenges faced by eSewa users?").

4. Research Models (Quantitative, Qualitative, Mixed Methods)

Research models determine how data is collected and analyzed.

a) Quantitative Research

Definition: Uses numerical data to test hypotheses. Characteristics:

• Objective and structured.
• Large sample sizes.
• Statistical analysis (regression, t-tests, ANOVA).

Example: Nepal Rastra Bank (NRB) conducts a study on inflation rates by collecting monthly price data from 5000 households and analyzing trends using regression analysis.

b) Qualitative Research

Definition: Uses non-numerical data (words, images, observations). Characteristics:

• Exploratory and flexible.
• Small sample sizes.
• Thematic analysis (interviews, focus groups).

Example: A WhatsApp Business team studies customer complaints by analyzing chat logs and conducting interviews with merchants to identify common issues.

c) Mixed-Methods Research

Definition: Combines quantitative and qualitative approaches. Characteristics:

• Triangulation (multiple methods for validation).
• Used when both numerical and descriptive data are needed.

Example: A Khalti study on digital payment adoption could:

1. Quantitative: Survey 10,000 users on payment frequency.
2. Qualitative: Interview 50 users on pain points in the app.

5. Selecting the Right Research Design

The choice depends on:

Decision Tree for Design Selection:

flowchart TD
    A["Start"] --> B{"Is the goal to explore?"}
    B -->|"Yes"| C["Exploratory Design"]
    B -->|"No"| D{"Is the goal descriptive?"}
    D -->|"Yes"| E["Descriptive Design"]
    D -->|"No"| F{"Is causality needed?"}
    F -->|"Yes"| G{"Can variables be controlled?"}
    G -->|"Yes"| H["True Experiment"]
    G -->|"No"| I["Quasi-Experiment"]
    F -->|"No"| J["Non-Experimental"]

6. Real-World Applications

a) Pathao’s Ride Pricing Model (Experimental Design)

Pathao uses A/B testing (a type of experimental design) to optimize ride prices. They:

1. Randomly assign two groups of drivers in Kathmandu.
2. Group A: Standard pricing.
3. Group B: Dynamic pricing (higher during peak hours).
4. Measure ride acceptance rates and driver earnings. Result: Dynamic pricing increased ride acceptance by 22% during peak hours.

b) Daraz’s Customer Satisfaction Survey (Descriptive & Quantitative)

Daraz conducts annual surveys to measure:

• Product quality ratings.
• Delivery speed satisfaction.
• Return rate trends. Method: Online questionnaire sent to 50,000 customers. Analysis: Uses descriptive statistics (mean, standard deviation) and regression to identify key drivers of satisfaction.

c) Ncell’s Network Optimization (Mixed-Methods)

Ncell wants to improve 4G coverage in rural areas. They:

1. Quantitative: Analyze call drop rates and signal strength data from 1000 towers.
2. Qualitative: Conduct focus groups with rural users to identify common complaints. Result: Identified obstacles like terrain and interference, leading to strategic tower placements.

d) NEPSE Stock Market Analysis (Non-Experimental)

Nepal Stock Exchange (NEPSE) uses time-series analysis (a non-experimental method) to:

• Predict market trends based on historical data.
• Study the impact of economic policies on stock prices. Example: After the 2023 budget announcement, NEPSE analyzed daily trading volumes to assess investor reactions.

7. Advantages and Disadvantages of Research Designs

8. Common Mistakes to Avoid

1. Choosing the wrong design: Using an experimental design when manipulation is impossible.
2. Ignoring ethical considerations: Random assignment may not be ethical in some studies (e.g., medical trials).
3. Poor operationalization of variables: Vague definitions lead to unreliable data.
4. Overlooking extraneous variables: Failing to control for confounding factors (e.g., weather affecting a survey).
5. Mismatch between research questions and design: A descriptive study cannot answer causal questions.

9. Worked Example: Traffic Congestion Study in Kathmandu

Research Problem: "What factors contribute to traffic congestion in Kathmandu’s Thapathali and Kalanki corridors?"

Research Questions:

1. What is the average daily traffic volume during peak hours?
2. How do public transport delays affect congestion?
3. Does one-way traffic implementation reduce travel time?

Research Design:

• Type: Mixed-Methods (quantitative + qualitative).
• Quantitative:
  • Data: GPS tracking of 1000 vehicles for 3 months.
  • Analysis: Regression analysis to identify key congestion factors.
• Qualitative:
  • Interviews: 50 drivers and commuters on perceived causes.
  • Observations: Traffic police logs on accident rates.

Expected Findings:

• Quantitative: Peak hours (7–9 AM, 5–7 PM) see 30% higher congestion.
• Qualitative: Drivers cite lack of signal synchronization and illegal parking as major issues.

Recommendations:

• One-way traffic in Thapathali could reduce travel time by 15%.
• Public transport priority lanes could ease congestion by 20%.

10. Exam Tip: How to Score Full Marks

Based on past exam patterns, here’s how to structure answers for Research Design:

a) For Short Notes (e.g., "Research Objectives")

• Definition: Start with a clear definition (e.g., "Research objectives are specific, measurable goals that guide the study.").
• Types: List 3–4 types (e.g., general, specific, primary, secondary).
• Example: Provide a real-world example (e.g., "A bank’s objective to ‘increase loan repayment by 10%’ is a specific, measurable goal.").
• Visual: Use a flowchart showing how objectives link to research questions.

b) For Long Answers (e.g., "Components of Research Proposal")

1. Introduction: Briefly define research design and its importance.
2. Components: Cover all key parts (research questions, hypotheses, variables, methodology, ethics).
3. Worked Example: Use a real scenario (e.g., "A study on eSewa’s transaction failures").
4. Diagram: Include a Mermaid flowchart of the research process.
5. Conclusion: Summarize how each component contributes to a valid study.

c) For Comparative Questions (e.g., "Quantitative vs. Qualitative")

Use a Markdown table with:

• Definition
• Data Type
• Methods Used
• Advantages/Disadvantages
• Example

d) For Problem-Solving Questions (e.g., "Which design for a bank study?")

1. Identify the problem: "Does SMS reminders improve loan repayment?"
2. Determine goals: "Test causality" → Experimental design.
3. Justify choice: "Random assignment ensures no bias."
4. Describe methodology: "Control group vs. treatment group."
5. Ethical considerations: "Informed consent from participants."

Final Checklist Before Submission

✅ Definitions: Clearly define all key terms. ✅ Examples: Use real-world Nepalese cases (e.g., Ncell, Daraz, banks). ✅ Visuals: Include at least 3 diagrams (flowcharts, tables, or labeled images). ✅ Structure: Follow logical flow (theory → examples → applications). ✅ Exam Focus: Align with past question patterns (components, comparisons, case studies).