This repository contains the handed in version of an end of term project in computational statistics. I took this course during the winter semester 2015/16 at Heidelberg University. The aim was to use tools developed in the lecture to analyse a simple data set. I decided to use the Geyser data set, which ships with every R installation.
I tried to build a framework in order to find good predictive models for the 'Old Faithful Geyser'. Therefore I introduced a skew cost function to account for the asymmetry in the problem (it is worse, if someone misses an eruption than if the person has to wait a little bit). With this I had to work out some fit routines as the standard regression tools do onlye support the standard, quadratic cost function. In the end I tried to measure the predictive capabilities of my prediction candidates.
The whole analysis was performed in R. The results can be found in ./Geyser-Analysis.pdf.