Is there a dependency between the weather and the number of traffic collisions?

Résumé du document

The major problematic of our assignment is trying to find if a relation between the weather and the number of traffic collision exists. In order to answer to this question, we had access to data from "MétéoSuisse" and "Police neuchâteloise". The weather data contains multiple variables such as temperature, humidity, precipitation and snowfall, while the traffic collision data contains the date and the number of accident recorded by the Police. This assignment contains different parts describing which statistic tools, databases, analysis and graphics we used.

There is a “Data” section, describing more specifically which variables we used and for what purpose. In that same section, there is a paragraph in which we explain the reason why we had to create a function and how we used it. There will also be a “Methodology” part explaining the way we analyzed a possible relation between weather condition and traffic collision.

The “Analyses” part of the project will show the results that came out of statistic tools such as, hypothesis test; correlation coefficient; graphics; data distribution. In the “Discussion” part, we will explain the way we interpreted the results, and verify if our intuition was right or wrong. The answer to our main question will be developed in this section. We included “R” codes that we used, in order to analyze our data, in section 8. Finally, we indicated sources we used in this project, such as our databases or some codes we used for the R software.

Extraits

[...] In Switzerland, and especially in the canton of Neuchâtel, we are aware of consequences of snow on the road, therefor we have machines and tools, to quickly move the snow away from roads and improve the driving security. By the way, this fact appears in our study. The only parameters we chose having an impact on traffic collision is the fresh snow. This parameter deals with the snow that has not been removed from the road, and this parameter has an impact on traffic collision. [...]

[...] The value Z is the result of difference between the two means corrected by the standard-error Thus, Finally, the last analysis instrument we will use to check the dependency between two variables is the correlation. Correlation can bring more information than hypothesis testing. Nevertheless, we must be aware of two facts. Correlations only measures linear relationship and do not always reveal a causality. We will calculate the rank correlation coefficient between each explanatory variables (weather) and the number of accident. It will return a value between 1]. [...]

[...] conditions” means conditions which could intuitively affect negatively the quality of the road, so the number of accident. That is to say low temperature, high humidity percentage, lot of snow or rain and high meteo coefficient. As we said, for each variable, we will separate the data into two groups (Bad and good conditions). The groups are separated by the median of the explanatory variable. The interest value is the mean (number of accident). The hypothesis we will test is the following: (Bad conditions # of accident mean)- (Good conditions # of accident mean) = 0 Hypothesis which we expect will come out false. [...]

[...] We will be using data of both cities. (To see the data in details, refer to the section annexes) Air humidity 2 meters above ground, daily mean, in percent 2. Rainfall, daily sum, in mm 3. Air temperature 2 meters above ground, daily mean, in Celsius degrees 4. Fresh snowfall, daily sum, in cm 5. Snow accumulation, morning measure, in cm We consider these indicators relevant because according to us, they are the main factors which could potentially influence the quality of the road. [...]

[...] Indeed, there was about 20'000 accident involving 312 death on the swiss roads in 2011. On the other side there were only 70 train accidents with 13 deaths in 2011, and 14 air transport accident with 13 deaths. It is obvious that car is one of the more dangerous means of. There is probably a lot of reason explaining these accidents. In our opinion, the weather plays an important role. Therefor we will exclusively focus on the impact of the weather on the number of accident. [...]