This versatile topic goes back to the inventions of Gauss, Markov, and Gibbs, whose ideas are incorporated in graphical models and regression graphs. Later, the geneticist S. Wright (1923– 1934) and the philosopher and computer scientist J. Pearl (1986– 1987) developed the tools, but their notation is too complicated to formulate the mathematical background. Here we mainly follow the up-to-date discussion of statisticians S. Lauritzen and N. Wermuth, and try to juxtapose the directed– undirected and discrete– continuous cases.