📊 A Hypothesis-Driven Analysis of Daily Spending Patterns (Weekdays vs. Weekends)
This project applies statistical hypothesis testing to understand whether personal spending behavior differs between weekdays and weekends. Using six months of credit-card transactions, the analysis explores a question I’ve been curious about:
Do I spend more money on weekends compared to weekdays — and is the difference large enough to influence budgeting decisions?
🧩 Problem Statement
The goal of this study is to test the assumption that people (myself included) tend to spend more during the week than on weekends. Using personal transaction data, I evaluate whether weekday spending is actually higher, using a combination of exploratory analysis and hypothesis testing.
🧪 Hypotheses
Null Hypothesis (H₀): Weekend median spending ≥ Weekday median spending
Alternative Hypothesis (H₁): Weekend median spending < Weekday median spending
📂 Data Description
This study uses seven months of personal credit-card transactions (Feb 1, 2025 → Sept 30, 2025) from two banks.
After cleaning and aggregating at the daily level, the dataset contains:
Metric Value Total Observations 209 daily records Features 4 (Date, Category, Type, Amount) Weekday Records 151 days Weekend Records 58 days 🔍 Methodology
- Exploratory Analysis
Summary statistics and distribution comparisons
Histograms + Q–Q plots showed right-skewness and high-value outliers
- Independence & Variance Checks
Durbin–Watson = 0.81 → Strong positive autocorrelation → Daily spending is not independent (typical for time-series data)
Levene’s Test → No significant variance difference → But due to skewness, non-parametric methods were used
- Hypothesis Testing Approach
Selected one-sided Mann–Whitney U test (non-parametric, robust to skewness and outliers)
Significance level: α = 0.05
📈 Key Findings
- Distribution Behavior
Weekdays: more transactions with lower dollar amounts
Weekends: fewer transactions but higher amounts
- Mann–Whitney U Test Results Statistic Value U 2561.0 p-value 0.999
Since p > 0.05, we fail to reject the null hypothesis.
🧠 Interpretation
A one-sided Mann–Whitney U test showed no statistical evidence that weekend spending is lower than weekday spending.
Instead, the results suggest:
Weekend spending is generally higher, not lower.
Thus the common intuition (“I spend more on weekends”) is supported by the data.
Data excludes seasonal effects (holidays, tax-season, etc.)
Transactions via cash, Venmo, or other sources are not included
Mann–Whitney compares distributions, not magnitude of overspending
💡 Recommendations
Based on the results:
Implement weekend-specific budgeting controls — weekend spending is the real driver.
Break down spending by category (e.g., dining, entertainment) to pinpoint high-impact areas.
Reduce weekend discretionary spending by 10–20% to produce meaningful annual savings.
Further category-level analysis will refine where overspending occurs.