tsFCI (time series Fast Causal Inference)

On this page you can find code of an adaption of the FCI (Fast Causal Infernce) algorithm (P. Spirtes, C. Glymour and R. Scheines (2000). Causation, Prediction, and Search. MIT Press, 2nd edition) to time series data, termed tsFCI. The method is introduced in the paper:

D. Entner, and P. O. Hoyer (2010). On causal discovery from time series data using FCI [pdf]
Proceedings of the 5th European Workshop on Probabilistic Graphical Models (PGM), Helsinki, Finland

Software

Below is the code for tsFCI (for Linux) including all commands to perform the simulations in the paper "On causal discovery from Time-Series Data using FCI" as well as calling tsFCI on a (real world) time series data set.
Note: This code is based on TETRAD version 4.3.9-18, see the TETRAD homepage.

Explanation	Download
To run the simulations of FCI, tsFCI, and Granger (the latter only in the infinite sample size). To run tsFCI on (real world) time series data sets.	R-code and (modified) TETRAD jar
To run the simulations of Granger on data, PSI and Group Lasso. Note: Granger causality was calculated using a free Matlab Toolbox downloadable from here. The code for PSI is available here. Finally, the code for Group Lasso was generously provided by Stefan Haufe, please contact him for that piece of code. (It will also need the SeDuMi optimization toolbox from Matlab downloadable from here.)	Matlab code
TETRAD code including all modifications for ts(C)FCI.	(modified) TETRAD source code

Contact Information

If you have any questions/comments, please contact Doris Entner.

Last update: 8 September 2010