Data Analysis & Report Writing: Tools, Techniques & Structure

Core Concepts & Definitions

1. Data Analysis: The Bridge Between Data and Meaning

Data analysis transforms raw data into actionable insights. It involves:

• Descriptive analysis: Summarizing data (mean, median, mode, standard deviation).
• Inferential analysis: Drawing conclusions about populations from samples (hypothesis testing, confidence intervals).
• Predictive analysis: Forecasting trends (regression, machine learning).

Why it matters: Without analysis, data is just numbers. For example, Nepal’s NTC collects call-drop data daily but only identifies network congestion hotspots after analyzing call logs and signal strength—leading to targeted tower upgrades.

2. Types of Data Analysis

3. Key Statistical Techniques

A. Descriptive Statistics: The First Step

Descriptive stats summarize data to reveal patterns. Common measures:

• Central tendency: Mean, median, mode.
• Dispersion: Range, variance, standard deviation.
• Shape: Skewness, kurtosis.

Worked Example: Daraz’s Order Fulfillment Time Daraz tracks delivery times (in hours) for 50 orders: [2, 3, 2, 4, 5, 3, 2, 6, 2, 3, ...]

1. Mean: $\frac{\sum X}{N}=\frac{150}{50}=3$ hours.
2. Median: Middle value (25th and 26th data points) = 3 hours.
3. Standard Deviation: Measures variability in delivery delays.
   • Use: Daraz uses this to set realistic delivery promises (e.g., "3±1 hours") and identify slow warehouses.

B. Inferential Statistics: Making Predictions

Used to test hypotheses and generalize findings. Key tests:

Mermaid Diagram: Hypothesis Testing Workflow

C. Data Visualization: Communicating Insights

A picture is worth 1,000 data points. Essential charts:

• Bar charts: Compare categories (e.g., NTC’s 3G vs. 4G coverage by district).
• Line graphs: Show trends over time (e.g., Nepal’s COVID-19 cases).
• Pie charts: Proportions (e.g., eSewa’s payment methods: 60% mobile, 30% bank, 10% cash).
• Scatter plots: Correlations (e.g., NEPSE vs. global oil prices).

4. Data Analysis Tools

5. Writing the Research Report

Reports follow the IMRaD structure (used in academic and corporate settings):

1. Introduction: Background, research gap, objectives.
2. Methodology: How data was collected/analyzed.
3. Results: Findings (use tables/charts).
4. Analysis: Interpret results (link to literature).
5. Discussion: Implications, limitations, recommendations.
6. Conclusion: Summary + future research.

Mermaid Diagram: IMRaD Structure

mindmap
  root((Research Report))
    Introduction["Problem statement\nLiterature review\nResearch questions"]
    Methodology["Design\nSampling\nData collection\nAnalysis tools"]
    Results["Tables\nCharts\nKey statistics"]
    Analysis["Interpret findings\nCompare with literature"]
    Discussion["Implications\nLimitations\nRecommendations"]
    Conclusion["Summary\nFuture work"]

A. Ethical Considerations in Reporting

• Plagiarism: Always cite sources (APA/MLA).
• Data integrity: No manipulation (e.g., Nepal Police cannot alter crime stats).
• Bias: Avoid leading questions in surveys (e.g., Pathao’s driver surveys must be neutral).
• Confidentiality: Protect participant data (e.g., Khalti anonymizes transaction records).

In the Real World

1. eSewa’s Fraud Detection:

   • Idea: Uses anomaly detection (a statistical technique) to flag unusual transaction patterns (e.g., sudden large payments from a new device).
   • How: Machine learning models analyze spending habits and alert users if behavior deviates from the norm (e.g., a usual $5/day user suddenly spends $500).

2. Nepal Rastra Bank’s Inflation Forecasting:

   • Idea: Time-series analysis (a predictive technique) to forecast inflation based on historical data (oil prices, import costs, monetary policy).
   • How: Economists run ARIMA models in R to predict inflation rates, which guide interest rate decisions.

3. Pathao’s Driver Satisfaction Surveys:

   • Idea: Mixed-methods analysis (quantitative surveys + qualitative interviews) to improve app features.
   • How: Pathao sends drivers a 5-point Likert scale survey (quant) and follows up with open-ended questions (qual). Thematic analysis reveals pain points like "low earnings in rainy season," leading to dynamic pricing adjustments.

Exam Tip

What Examiners Look For

1. Technical Accuracy:

   • Know when to use t-tests vs. ANOVA (e.g., t-test for 2 groups, ANOVA for 3+).
   • Example question: "A study compares internet speeds in Kathmandu, Pokhara, and Chitwan. Which test is appropriate?" → ANOVA.

2. Report Structure:

   • Marks are deducted for missing IMRaD sections. Always include:
     • A clear research question in the intro.
     • Statistical justification (e.g., "We used a chi-square test because both variables were categorical").
     • Visuals with captions (e.g., "Figure 1: Distribution of Khalti users by age group").

3. Real-World Application:

   • Exams often ask: "How would [company X] apply [technique Y]?"
   • Example answer:

     "Ncell could use cluster analysis to segment customers by usage patterns (e.g., heavy data users vs. casual callers) and offer targeted plans. This is similar to how Pathao uses clustering to group drivers by efficiency and assigns them high-demand routes."

4. Common Pitfalls:

   • Ignoring assumptions: Always state assumptions (e.g., "ANOVA assumes homogeneity of variance").
   • Overinterpreting p-values: A p-value of 0.06 is not statistically significant (cutoff is 0.05).
   • Poor visuals: A bar chart with no labels or a table without a caption loses marks.

Practice Question with Solution

Question: "A researcher wants to analyze whether students who use Khalti for tuition payments have higher academic performance than those who use bank transfers. Design a study and choose the appropriate statistical test."

Solution:

1. Research Design:

   • Type: Quasi-experimental (no random assignment; groups are pre-existing).
   • Variables:
     • Independent: Payment method (Khalti vs. bank transfer).
     • Dependent: Academic performance (GPA).
   • Sampling: Stratified random sampling (e.g., 100 Khalti users, 100 bank transfer users from TU).

2. Data Collection:

   • Survey students on payment method.
   • Obtain GPAs from university records.

3. Statistical Test:

   • Independent samples t-test (since we compare means of two independent groups).
   • Assumptions:
     • Normality (check with Shapiro-Wilk test).
     • Homogeneity of variance (Levene’s test).

4. Hypotheses:

   • H₀: There is no difference in GPA between Khalti and bank transfer users.
   • H₁: Khalti users have higher GPAs.

5. Reporting:

   • Results section:

     "An independent samples t-test revealed a significant difference in GPAs (t(198) = 2.34, p = 0.02), with Khalti users (M = 3.2) outperforming bank transfer users (M = 3.0)."

   • Discussion:

     "This aligns with prior research on digital payment efficiency reducing administrative delays (e.g., eSewa’s impact on bill payments). However, further qualitative data (e.g., interviews) could explore whether Khalti users are inherently more tech-savvy."