Methodology

Data Collection

The initial dataset was sourced from two Wikipedia pages that list all the German U-Boats from U-1 to U-4712. The data collection process involved:

1. Primary Sources: Wikipedia pages listing German U-Boats
2. Secondary Collection: Individual U-Boat pages for commissioning dates
3. Data Fields: U-Boat name, year, type, notable commanders, damage statistics, fate, and notes

Data Processing Pipeline

Step 1: Web Scraping

• Tool: Python with BeautifulSoup and requests
• Process: Automated scraping of U-Boat data from Wikipedia
• Output: Raw CSV files with U-Boat information

Step 2: Data Cleaning

• Tool: R with tidyverse
• Process:
  • Removed unnecessary columns
  • Calculated active service days (commissioning to fate date)
  • Standardized categorical variables
• Output: Clean dataset ready for analysis

Step 3: Survival Analysis

• Tools: R with survival and survminer packages
• Methods:
  • Kaplan-Meier survival curves
  • Cox Proportional Hazards modeling
  • Log-rank tests for group comparisons

Statistical Methods

Survival Analysis Approach

We treated U-Boat “death” (sinking, capture, scuttling) as events and calculated time-to-event from commissioning date.

Key Metrics

• Active Service Days: Primary outcome variable
• Event Status: Binary indicator of U-Boat fate
• Covariates: U-Boat type, commissioning year, notable commanders

Model Validation

• Proportional hazards assumption testing
• Residual analysis
• Cross-validation techniques

Limitations

1. Data Quality: Reliance on Wikipedia may introduce inaccuracies
2. Missing Data: Some commissioning dates were unavailable
3. Survival Bias: Focus only on documented U-Boats
4. Historical Context: Limited accounting for operational changes during war

Back to Main Analysis