Survival on the Titanic
by XEM Team
Summary
Titanic, said to be “unsinkable”, sank in 1912, killing 1502 out of 2224 passengers. One of the first big accidents with data about it which was also been used to improve maritime safety by passing new policies, better procedures and construction to avoid similar catastrophes.
We analysed original data from the accident and explored how different attributes of passengers are related to their survival. It must be noted that we have data on 891 passengers out of 2224 people on board and we don’t know how representative our sample is out of the whole data. These attributes include gender, age and wealth. We used logistic regression model to try and predict survival based on these attributes. Lastly, we created a mosaic plot on the basis of a historic plot, short time following the accident, to display the relationship between different variables and survival.
Age and survival
From the dataset, age is given in years. We expect to see a relationship between age and survival based on different age groups rather than the numerical age itself, so we decided to convert ages into categorical groups ( children, teenagers, young , middle aged and elderly adults). This allowed us to create a barplot to show how the age of passengers and their survival are related. We found that children have higher survival rate than the rest age groups and survival rate generally decreases for higher age groups. This can be explained by the fact that it’s in human nature to give priority to saving infants and children, this was indeed reported as order from the captain.
Gender and survival
During these times, in the spirit of chivalry, women were saved first. This was historically recorded as orders from the captain of the Titanic. Based on this, we wanted to verify if gender affected the survival rate. Since this is historical data, gender is only recorded as male and female, thus we used binary data for our research. A barplot was created, using gender as the predictor variable and percentage of survival displayed in terms of colours showing survivals or not as the outcome variable. The graph showed that the percentage of survival of female and male was 74.2% and 18.9%, respectively. We concluded that female had higher survival rate than male.
Wealth and survival
Ticket Price and Passengers’ class are major indicators of socio-economic status of passengers. By displaying ticket price distribution by class, using boxplots, shows that 1st Class has the highest median ticket price for both survivals and non. Following that, by finding the percentage of survival by class, using barplot, it is noted that the percentage of survival for 1st class is more than double than the percentage of the 3rd. Port of Embarkation has been also considered as an indication of passenger’s wealth. By finding survival rate by class along with median ticket price, per port, it is shown that there is an association between port of embarkation, ticket price, passengers’ class and survival rate.
Model
We fitted a linear regression model to our data, trying to use the different variables available to us from the dataset and compared the predicted properties. We focused on making our model parsimonious and saw that the explanatory variables, which allow for the model with better predicted performance are indeed gender, age and passenger class. We then used and introduced different statistical tests (not in the syllabus), as well as a ROC course to display and evaluate the model’s strength on the test data. The model predicts that young, 1st class, female passengers are more likely predicted to survive compared to the counterparts of each of the attributes. It’s worthwhile noting that the purpose of a predictive model when it refers to historic data like this is rather than predict per se, provide an overall better understanding of the data and what it refers to.
Recration of historic plot
One of the first visualizations that had ever being created with data from the Titanic is a graph by graphic artist G. Bron, published on Sphere (British newspaper) one week after the accident. His work is an early innovation in data display where each subgroup shown by a bar with area proportional to the numbers of cases which today can be seen as an early mosaic plot. We decided to work on our own mosaic plot to display how all different factors analysed are related to survival in a single visualization.
Presentation
Our presentation video can be found here.
Data
CKAN training (2014). Titanic [Dataset], viewed 22 October 2021, https://data.wu.ac.at/schema/datahub_io/MWViYjhmYTctMzE2NS00OWEwLThmZDgtMTUwZjI4MThiYTJl.
References
[1] Source: Friendly, M., Symanzik, J., & Onder, O. (2019, February 6). Royal Statistical Society Publications. Royal Statistical Society. Retrieved November 19, 2021, from https://rss.onlinelibrary.wiley.com/doi/full/10.1111/j.1740-9713.2019.01229.x.
https://rkabacoff.github.io/datavis/Models.html#Mosaic
https://www.houstoninjurylawyer.com/titanic-changed-maritime-law/
https://www.datavis.ca/papers/titanic/