„Improving statistical inference in big data sets with spatiotemporal structure” wird im Rahmen der Förderlinie E-Club: Anschubfinanzierung für Brückenprojekte durch die Freie Universität gefördert. Das Projekt beschäftigt sich mit der Schnittstelle von theoretischer Statistik, der Analyse von FMRT-Daten (funktionelle Magnetresonanztomographie), und effizientem Algorithmen-Design in parallelen Rechenarchitekturen.

 

Leitung: Prof. Dr. Dirk Ostwald (Center for Cognitive Neuroscience Berlin) und
              Prof. Dr. Timo Schmid (Institut für Statistik und Ökonometrie)

Laufzeit: 1.7.2015 – 31.3.2016

Förderungssumme: 14.800 €

 

Project description: The technological advances of the digital revolution have enabled researchers to acquire large digital data sets (“Big Data”) with relative ease. However, most statistical techniques that are used to extract meaning from Big Data were developed with much smaller data sets in mind. These classical techniques thus only make suboptimal use of the rich spatiotemporal structure that Big Data can exhibit. The aim of the current project is to build on recent advances in theoretical statistics known as "Empirical Bayes" to develop a novel statistical method for the analysis of large data sets with deep spatiotemporal structure. Specifically, we will address a recent trend in cognitive neuroscience to acquire functional neuroimaging data from very large cohorts of participants (N=100 to N = 1000), for which no adequate inference method exists so far.