Somogyvári Zoltán (Wigner RMI): Determination of temporal structure of causal effects in time series
→
Europe/Budapest
Tanacsterem
Tanacsterem
Description
Identification of causal relationships between time series is one of
the most important question in many branches of science. The core of
this question is that how we can distinguish between common cause,
unidirectional direct causal effects or the circular causality which
can not be assessed by the different forms of correlation. The most
widely used and Nobel-price awarded method, the Granger-causality uses
linear regression models, assumes weak connections and become
ambiguous in case of circular causality. George Sugihara et al. in
2012 introduced a new framework for causality analysis, using the
phase-space reconstruction based on the Takens-theorem. It was
demonstrated that the new method is be able to distinguish between the
correlation and causality even in the circular case however the new
method provides information only on the existence of a causal
relation. We extended the Sugihara method in order to identify
temporal delays in causal effects and made it applicable for electric
potential recordings in the brain.
In the talk, we present the background of causality analysis and the
new method, as well as results on simulated coupled chaotic systems.
