What is it about?
We propose a framework and its theoretical aspects for inferring causal relations of time series based on the concept of Variable-lag Granger causality and Transfer entropy. Given two time series, we want to know which time series influence/causes another time series values. The framework is included in VLTimeCausality R CRAN package.
Featured Image
Photo by Agê Barros on Unsplash
Why is it important?
Granger causality and Transfer Entropy are widely used for inferring causal relations between time series. However, these models assume that every point in the time series is influenced with a fixed time delay. This is not true in real world situations (e.g. trajectories of walking and suddenly running when following someone). In this work, we relax this fixed-lag assumption and provide a theoretical framework for Granger causality and Transfer Entropy estimation with varying delay lags between time series.
Perspectives
I really like the paper. As an author, it was fantastic to have a chance to think and to generalize the old models like Granger causality and Transfer Entropy that people used for very long time. I feel like our generalization can pave the way for better understanding of causal relations in nature.
Dr. Chainarong Amornbunchornvej
National Electronics and Computer Technology Center
Read the Original
This page is a summary of: Variable-lag Granger Causality and Transfer Entropy for Time Series Analysis, ACM Transactions on Knowledge Discovery from Data, August 2021, ACM (Association for Computing Machinery),
DOI: 10.1145/3441452.
You can read the full text:
Resources
Contributors
The following have contributed to this page
