FIGURE 1.

Identifying causal effects in neuroimaging time series using the arrow‐of‐time. (a) Since cause precedes effect, causal effects in multivariate time series cannot be identified from metrics that are blind to the AoT. Such symmetric metrics, for example, mean or average correlation over time points, are equal in forward and backward data. In contrast, asymmetric metrics are different in forward and backward data as they are sensitive to the arrow‐of‐time, thereby bearing the potential of capturing causal effects. (b) We use fMRI time series acquired during resting state and seven different tasks. The AoT signature is evaluated in these time series using Equation (2), and the amplitude of the causal effect is assessed by comparison against null time series with no causal effects.