LIP External Colloquium: Steven Boker, University of Virginia - Path Diagrams and Function Nodes: A Proposed Extension to SEM

Steven Boker, University of Virginia

Path Diagrams and Function Nodes: A Proposed Extension to SEM
Path diagrams, and in particular RAM style path diagrams, express a relationship between graphics and mathematics that helps us understand complicated linear models for the associations between measured and unmeasured variables. In the abstract, a path diagram is an organization of nodes and edges. The OpenMx team has been discussing extending SEM and related path diagrams in a variety of ways in order to allow a more broadly understandable representation of nonlinear models. One proposal is for what we call Function Nodes. One may think of so-called "dummy variables" used in OpenMx as Addition Nodes, so this proposal is an extension of current practice. Function Nodes would have defined inputs and outputs, i.e., domain and range, that would allow pre-defined functions to be computed as part of the path diagram. Some example Function Nodes will be provided along with reasoning behind their use. In this seminar I hope to engage in a discussion of how to improve the proposal and to elicit alternative ideas about how to bring nonlinear modeling into the path diagrammatic framework. For instance, are there useful alternatives to the current use of one headed and two headed arrows as the edges in a path diagram? Should we think about encapsulizing sections of a path diagram to make these sections reusable? We are interested in suggestions that are feasible, improve the user's understanding, and enable new ways of thinking about and modeling.

Join on-site or hybrid:

https://mpib-berlin.webex.com/mpib-berlin/j.php?MTID=m314575b44cf8265f4af1ffa9188fe2dd

Meeting number: 2744 871 2321

Password: m9qPmEPdq33