Graphical models have been developed as an extension of path analysis to general forms of dependencies generated over graphs. An example is given of a graph derived from a set of data. It illustrates how the graphical summary visualizes important aspects of analysis and helps to communicate main results.
It is outlined when and why results for linear systems extend to distributions generated over the same type of graph. It is shown how they can aid in separating possible confounding effects from direct dependencies and in studying causal
hypotheses.