Dresden 2006 – scientific programme
Parts | Days | Selection | Search | Downloads | Help
AKSOE: Physik sozio-ökonomischer Systeme
AKSOE 10: Poster Session (posters are expected to be displayed the full day 8:30-18:00)
AKSOE 10.46: Poster
Wednesday, March 29, 2006, 16:00–18:00, P2
Filtering tools for correlation networks — •Jan Nagler1,2, Philipp Weber2, Shlomo Havlin2,3, and H. Eugene Stanley2 — 1Institute for Theoretical Physics, Bremen University, Bremen, Germany — 2Center for Polymer Studies, Boston University, Boston, USA — 3Minerva Center and Dept of Physics, Bar-Ilan University, Ramat-Gan, Israel
We present a novel method for filtering information from cross-correlation data. The key point of our method is a reconstruction process of the original cross-correlation matrix from a filtered correlation network. In order to recover the original correlation matrix we theoretically derive a procedure capable to estimate unknown correlation coefficients from known ones. We apply the method to correlation networks of stock market returns. First, we compare clustering with standard stock sector classification schemes. Second, we check whether our proposed procedure is taylored to evaluate other filtering methods (e.g. Minimal Spanning Tree based methods or approaches applying percolation schemes).