From temporal data to dynamic causal models

We present a brief review of dynamic causal model inference from data. A vector autoregressive models is our prime interest. The architecture, representation and schemes measurement temporal data time series are outlined. argue that require- ment to characteristics should come the nature process at hand goals inference. To describe evaluate one may use terms longitude, frequency etc. Data crucial factor in order an inferred be adequate. longitude observation session duration expressed via several horizons, such as closest horizon, 2-step influence attainability oscillatory evolutionary horizon. justify data, analyst needs assume stationary or least obeys structural regularity. main specificity task known variables certain If maximal lag unknown, faces additional problems. examine Granger’s causality concept outline its deficiency real circumstances. It argued Granger incorrect practical tool discovery. In contrast, rules edge orientation (included constraint-based algorithms inference) can reveal unconfounded relationship.