Phone: +49 6131 39 20141
Applying multivariate statistics for hypothesis testing and data exploration in experimental animal studies
Small sample size in animal research poses a major challenge for the statistical analysis of data from preclinical studies. Depending on the type of data (e.g. omics studies), sometimes the number of outcome measures is much greater than the number of measurements per variable (“great p, small n “ problem). Analysis of such data requires methods which are able to deal with their multi-dimensional nature. While principal component analysis (PCA) has emerged as the most popular dimensionality reduction technique, numerous other methods might be more appropriate depending on the type of data and research question. In my thesis, I am investigating which ordination techniques lead to robust inferences in data sets with more variables than measurements. Furthermore, I aim at identifying which methods offer increased sensitivity to detect group differences. Beyond the theoretical comparison of multivariate methods using simulated data, I also apply ordination techniques for the analysis of omics data from animal studies including for example 16S sequencing or lipidomics data.