An end-to-end data science assignment utilizing programmatic statistical testing to analyze online retail transaction data and evaluate standard hypothesis tests.
This project explores the relationship between e-commerce transaction attributes (such as quantity, unit price, and geography) and purchasing behavior. By applying practical Python-based statistical testing to large-scale data, this project demonstrates a structured approach to analyzing highly skewed financial datasets and validating fundamental statistical assumptions.
Statistical Testing on E-Commerce Dataset.ipynb: Jupyter notebook containing the full data pipeline, EDA, and programmatic hypothesis tests.data.csv: The primary dataset used for the analysis. https://www.kaggle.com/datasets/carriel/ecommerce-dataStatistical Testing Report.pdf: The final LaTeX-compiled academic summary of the findings.
- Exploratory Data Analysis (EDA): Performed data cleaning, feature engineering (
revenue), and visualization (histograms, boxplots) to identify severe right-skewness and extreme outliers in e-commerce financial metrics. - Statistical Testing (Python): Implemented Shapiro-Wilk, Spearman rank correlation, Welch's two-sample t-test, One-Way ANOVA, and Chi-Square tests using
scipy.statsto segment and analyze transactional behavior across different countries.
- Python: Data manipulation and programmatic testing.
- SciPy/Pandas: Statistical functions, hypothesis testing, and dataframe management.
- Seaborn/Matplotlib: Statistical data visualization.
- LaTeX: Academic report typesetting.
| Metric | Value |
|---|---|
| Revenue Normality (Shapiro-Wilk p-value) | < 0.0001 (Assumption Violated) |
| Price vs. Qty Correlation (Spearman |
-0.4032 (Inverse Relationship) |
| UK vs. Rest of World (t-statistic) | -30.8258 |
| Top Countries Variance (ANOVA F-Statistic) | 10.2228 |
| High-Value Independence (Chi-Square Stat) | 358.1028 |
Sepehr Barekati