Statistical Data Analysis
The statistical data analysis is used to prepare and present data to test hypotheses and to derive recommendations for design and/or actions. We offer support during all stages of the analysis process. These include …
- … the selection of appropriate descriptive and inferential statistical methods depending on the research questions, type and amount of the available data,
- … data preparation, e.g. checking for consistency, data aggregations, and gap analyses.
- … the implementation of selected methods: e.g. t-test, variance and correlation analyses and structural equation modeling,
- … and the interpretation of results, evaluation of their validity in relation to their statistical significance, their practical significance and methodical limitations.
The research report will be written in accordance with scientific standards if desired.
Various problems can be examined using machine learning methods. For example, regularities, which are not detectable in the available experimental data by standard statistical procedures, can be found (“unsupervised machine learning”). Also descriptive models can be created for further investigation using these identified relationships. Moreover, multidimensional functions for the separation of different classes can be generated using existing foreknowledge – e.g. with using knowledge about class memberships. With the help of these hyperplanes predictive models can be generated in order to predict class memberships for unknown data sets (“supervised machine learning”).
A typical approach might be as follows: the most relevant features are extracted from the existing experimental data, which are reduced to the ones relevant for classification using feature extraction methods. One or more machine learning algorithms (e.g. support vector machines and artificial neural networks) generate models using the so created feature sets with which the corresponding question can be further investigated.