Plane Crash Fatalities Analysis

Overview

This project analyses plane crash fatalities over the past century (1918-2022) using data from Kaggle. The aim is to explore factors affecting plane crash outcomes, focusing on fatalities, flight type, crash causes, and regional effects.

Research Questions

The analysis seeks to answer the following:

1. Flight Type & Fatalities: Is there a significant difference in the number of fatalities between the most frequent two flight types?
2. Flight Type & Proportion of Fatal Flights: Do different flight types vary in their likelihood of resulting in fatalities?
3. Crash Cause & Fatalities: Does the cause of the crash significantly affect the likelihood of fatalities?
4. Region & Fatalities: How does the region affect the number of fatalities in plane crashes?

Data Preparation

• Dataset: Cleaned by removing missing values and selecting relevant columns.
• Derived Variables: Created a binary indicator (has_fatalities) to track whether a crash led to fatalities.
• Sampling: A random subset of 1000 records was taken for computational efficiency.

Key Findings

1. Flight Type & Fatalities

• A t-test revealed that Commercial flights tend to have significantly higher fatalities than Operational flights.
• P-value: 9.4e-08 → Strong statistical significance.
• Confidence Interval: Does not include 0, reinforcing a meaningful difference.
• Visuals: A boxplot showed greater variability in Commercial flight fatalities, with notable outliers.

2. Flight Type & Proportion of Fatal Flights

• A proportion test found no statistically significant difference in the likelihood of a crash resulting in fatalities between Commercial and Operational flights.
• P-value: 1 → We fail to reject the null hypothesis.
• Conclusion: The number of fatalities may differ, but the probability of a flight resulting in fatalities is similar between the two flight types.

3. Crash Cause & Fatalities

• ANOVA test showed that crash cause significantly affects the occurrence of fatalities.
• P-value: Very low, indicating strong statistical significance.
• F-value: 9.712 → Crash cause plays a substantial role in determining fatalities.
• Visuals:
  • Bar plot: Highlights the most fatal crash causes (e.g., terrorism has the highest fatality rate).
  • Mosaic plot: Shows how different crash causes correlate with fatality likelihood.

4. Region & Fatalities

• Linear regression was used to examine the effect of region on the number of fatalities.
• Findings:
  • Low R-squared: Indicates region is not a strong predictor of fatalities.
  • F-statistic: 1.456 → Weak explanatory power.
  • Bar plot: Displays regions with the highest total fatalities.
• Conclusion: Other factors likely play a more significant role than geographic location in determining crash fatalities.

Technologies Used

• Language: R
• Libraries: dplyr, ggplot2, base R functions
• Statistical Methods: t-tests, proportion tests, ANOVA, linear regression
• Data Visualisation: Boxplots, bar plots, mosaic plots