DAG thinking (student stuff)

21 Jan, 2022 at 10:30 | Posted in Statistics & Econometrics | 2 Comments



  1. For everyday life my first language is English, which is an excellent language for most purposes.
    I know enough French and German to know that these are also excellent languages, but I rarely need to use them.
    I don’t know or need Chinese, Japanese, Tamil etc but there is no doubt that these are also excellent languages.
    For specialised statistical work, like most economists, my first “language” is frequentist, which is hard but effective.
    I am attracted to and havea limited knowledge of Bayesian statistics.
    Both of these languages commonly make use of flow diagrams.
    I have also struggled to gain familiarity with a strange new language called “do calculus” or “path analysis” associated with Judea Pearl and his acolytes.
    This language starts with a “Directed Acyclic Graphs” (DAGs) which are merely a limited (Acyclical) type of flow diagram.
    A variety of peculiar terms are then used (“forks”, “backdoors”, colliders”, “d-separation” etc).
    Soon we are talking a weird new language which scrambles your brain.
    What can be gained from mastering this strange language?
    Zilch !
    Existing statistical languages are far more powerful and much easier to understand.

    • From wikipedia:
      《Electronic circuit schematics either on paper or in a database are a form of directed acyclic graphs using instances or components to form a directed reference to a lower level component. Electronic circuits themselves are not necessarily acyclic or directed.》
      When you represent natural systems such as language in a directed graph, are cycles a feature not a bug?
      Maybe you lose some closures, but do you gain in expressivity by allowing cycles? In the video example, might not parent learn from child?

Sorry, the comment form is closed at this time.

Blog at WordPress.com.
Entries and Comments feeds.