Statistical Tests



Any statistical analysis is worthless if its conclusions are the result of random variance or any other kind of statistical error. It is therefore of essential importance to check the experiment and its results as strictly as possible. Statistical test are a very helpful for this purpose. Our framework provides about 30 different statistical tests which can be executed directly by using the corresponding Java class and checking significance or correlation. The correctness of calculation is assured since the current implementations are using R for execution. The results are returned in a public hash table and in an EngineResult object. This allows fast and direct access as well as further utilization in other engines or for visualizations. Furthermore the test classes contain an object orientated description vector describing their calling parameters. This information is used to generate the test dialogs according to the chosen test in the Bioinformatic Studio.