Our research focuses on causal inference for multidimensional data, including time series data, centered around a new method called Volatility Constrained Correlation (VC correlation).
In various fields, efforts have been made to estimate the underlying network by computing correlation matrices from multidimensional data, including time series data. These studies typically use the Pearson product-moment correlation coefficient to identify networks from multidimensional data. However, while regular correlation coefficients can indicate the presence or absence of correlation, they do not reveal the directionality of the correlation. In other words, for a pair of data A and B, the correlation coefficients Cor(A, B) and Cor(B, A) yield the same value, making it unclear which of A or B is the controlling side and which is the controlled side.
Therefore, we have developed a new type of correlation estimation method called Volatility Constrained Correlation (VC correlation) [Paper 1]. By using VC correlation, it is now possible to estimate not only the correlation between two elements, A and B, but also the direction of causality.
We are conducting research applying VC correlation to financial data [Paper 1] [Paper 3] [Paper 4], to biomedical data [Paper 2] and to elucidate the theoretical structure of VC correlation.
papers related to VC correlation
[Paper 1] T. Ochiai, J.C. Nacher, “Volatility-constrained correlation identifies the directionality of the influence between Japan’s Nikkei 225 and other financial markets”,
Physica A 393, 364–375 (2014)
(Link to the paper)
[Paper２] T. Ochiai, J.C. Nacher, “Predicting link directionality in gene regulation from gene expression profiles using volatility-constrained correlation”,
Biosystems, Volume 145, 9–18 (2016)
(Link to the paper)
[Paper３] T. Ochiai, J.C. Nacher, “VC correlation analysis on the overnight and daytime return in Japanese stock market”
Physica A, Volume 515, 537-545(2019)
(Link to the paper)
[Paper４] T. Ochiai, J. C. Nacher, “Unveiling the directional network behind financial statements data using volatility constraint correlation”,
Physica A: Volume 600, 127534, 12 pages (2022)
(Link to the paper),(PDF)