RAND

RAND is a repository for a sport dataset analysis made with the language R. The analysis is done in Projet_final_Lemonnier_Alexandre_Simonin_Victor.ipynb with a dataset description, some statistical analysis, visualisation, a PCA (Principal Component Analysis ) and finally a logistic regression.

Dataset

The dataset is here in /data/decathlon.csv. The dataset is made up of several quite different pieces of data. These represent the scores obtained by athletes in the decathlon event at the Olympic Games and at the Decastar, we have the scores of the 10 events as well as their ranking and their points.

PCA

We perform a PCA (Principal Component Analysis) in the notebook.

This allows us to analyze and visualize our Decathlon dataset, which contains individuals described by several quantitative variables.

PCA is a method that allows us to explore data with multiple variables. Each variable could be considered as a different dimension. This is useful because it could be very difficult to visualize our data in a multidimensional "hyper-space".

Correlation

We also visualize our data with a correlation graph :

We made few analysis on it:

• The results of the short race events, therefore 100m, 400m and 110m.H, are strongly correlated with each other. Quite surprisingly, however, they are negatively correlated with the final ranking, unlike the Length, Weight and Height events.
• The Length event and the 400m event are quite strongly negatively correlated.

Representation

Finally there is a representation on the factorial axis that allows us to notice the quality of representation of our variables. Here, the axes must be interpreted independently.

Several elements can be analysed:

• Positively correlated variables are grouped together.
• Negatively correlated variables are positioned on opposite sides of the origin of the graph (opposing quadrants).
• The distance between the variables and the origin measures the quality of representation of the variables. Variables that are far from the origin are well represented by PCA. For example, the throwing events, namely Shot Put, Discus and Javelin, are strongly correlated because they point in the same direction. However, Javelin is less well represented than Weights and Discs in the first two principal components.