gma-package {gma}R Documentation

Granger Mediation Analysis

Description

gma performs Granger Mediation Analysis (GMA) of time series. This package includes a single level GMA model and a two-level GMA model, for time series with hierarchically nested structure. The single level GMA model for the time series of a single participant performs the causal mediation analysis which integrates the structural equation modeling and the Granger causality frameworks. A multivariate autoregressive model of order p is employed to account for the spatiotemporal dependencies in the data. Meanwhile, the model introduces the unmeasured confounding effect through a nonzero correlation parameter. Under the two-level model, by leveraging the variabilities across participants, the parameters are identifiable and consistently estimated using a two-stage method or a block coordinate descent method.

Details

Package: gma
Type: Package
Version: 1.0
Date: 2017-08-23
License: GPL (>=2)

Author(s)

Yi Zhao <zhaoyi1026@gmail.com> and Xi Luo <xi.rossi.luo@gmail.com>

Maintainer: Yi Zhao <zhaoyi1026@gmail.com>

References

Zhao, Y., & Luo, X. (2017). Granger Mediation Analysis of Multiple Time Series with an Application to fMRI. arXiv preprint arXiv:1709.05328.


[Package gma version 1.0 Index]