This is a collaborative space. In order to contribute, send an email to maximilien.chaumon@icm-institute.org
On any page, type the letter L on your keyboard to add a "Label" to the page, which will make search easier.
Network causality
Causality Analysis toolbox V0.3
This document is meant to provide some principal guidelines to the user of this toolbox, in order to facilitate his/her acquaintance with the available tools.
The main functionalities that this toolbox provides are summarized below:
This toolbox can be used to perform data-driven causality analysis (implemented methods: PDC, GPDC, DTF, DC, dDTF, dDC and PGC) on a provided set of input time series.
Both the parametric (Multi-Variate AutoRegressive modeling) approach and the non-parametric (multi-taper spectral estimation, followed by a spectral decomposition step, as initiated by Dhamala et al, 2008) approach are implemented. The non-parametric approach is actually implemented, also in combination with integrated time-frequency analysis (wavelet spectral estimation, followed by a spectral decomposition step, as initiated by Dhamala et al., 2008), interesting to be applied in the case of highly non-stationary (rapidly time-varying) causal effects.
Several preprocessing steps are possible to be applied to the input time series, before the causality analysis (downsampling, detrending based on a linear model, temporal demeaning, ensemble normalization in the case of multitrial data, differencing).
The phase randomization method is implemented for the generation of surrogate distributions and the significance thresholding of the causality results. The user is called to define the number of phase randomizations to be performed, the significance threshold 'alpha' and the type of correction for multiple comparisons (three options: NO correction/Bonferroni correction/ False-Discovery-Rate (FDR) correction).
Two modes of visualization of the causality results are provided:
a basic visualization in the form of an [Nvar x Nvar] array of figures for each causality measure, Nvar being the number of network nodes (= number of input time series). In this array, each figure provides either a spectral or a time-frequency representation of the causality from the 'column' to the 'row' network node.
- A spatial visualization of the causality on the cortex (applicable in the case of source causality analysis), using a slightly adapted version of routines found in the open-source software package eConnectome.
The causality toolbox is available in /usr/global/matlab/mega_toolbox/mega_causality. Add this path to you matlab path to use the toolbox.
The complete documentation V 0.2 is available in /usr/global/matlab/mega_toolbox/mega_causality/Documentation or here :user_s_guide_causality_analysis_toolobox_v0.2.pdf|User Guide
The release notes for V 0.3 are available here realase notes
The corresponding paper is available here :comparative_performance_evaluation_of_data-driven_causality_measures_applied_to_brain_networks.pdf|JNeuMeth Paper