Structural causal modelling is a framework for developing causal hypotheses to test with actual data. You can look up the strict mathematical formulation of it, though I'll advocate here more for a practical approach:

As for the test, one if its terrific features is that you can use any test you want - generalized linear models for example.

You can apply it quite readily in R - there are packages like "daggity" or "ggdag" available.

I am particularly grateful for the information you shared about using generalized linear models and the R packages "daggity" and "ggdag" for applying Structural Causal Modelling in practical scenarios.

What intrigues me the most about Structural Causal Modelling is its ability to unveil causal relationships among variables, especially in complex systems. In a recent paper titled "Causal Analysis of an Agent-Based Model of Human Behaviour" by Marcel Kvassay and colleagues, the authors utilized the concept of a causal partition of a model variable. This approach was employed as predictors of emergence, providing a fascinating perspective on the topic. While I found the article compelling, it is my first encounter with the Structural Causal Modelling approach, and I find it challenging to grasp fully.

Warm regards,

For example, structural causal modeling can be used to define the evolution function that will be applied in the definition of a cellular automaton or agent-based model of a complex system.

From my own experience, the finding of the as precise as possible evolution function that is used in the particular CA model of a complex system is the crucial part of each model definition.

Hence, initially, structural causal modeling is applied to reveal the information flow through the system, which is then fed into a CA model. The model is tested and compared to the modeled phenomenon. The process is repeated so many times until it satisfactory reproduce the observed phenomenon. It can be done in the similar manner in other types of CS models.