Efficient Dimension Reduction using Dynamic Functional Principal Components

By Siegfried Hörmann With increasing complexity and the rapidly growing amount of data collected in almost all areas of our life, it becomes more difficult to draw meaningful conclusions and to filter relevant information. The field of statistics has seen a big upsurge due to such new challenges. My research is devoted to some of these challenges. Over the past decades storing and collecting data electronically has steadily become easier and cheaper. As a consequence, for many everyday life processes or scientific experiments nearly continuous data records exist. For example, on some engine test benches hundreds of variables can be collected and it is not uncommon to have for certain parameters of interest several measurement points per second. Similar examples can be given in environmental sciences (pollution levels), geophysics (strength of magnetic fields), medicine (fMRI images) or econometrics (tick-data), to just name a few.