Time complexity of constructing cause-and-effect diagram
oceanzwq opened this issue · comments
Hello，
Dowhy is an effective tool for building cause-and-effect diagrams, so what is the time complexity of building a causal diagram? Can multiple causal diagrams be provided based on existing data and inputs? What are the consequences if a causal diagram is constructed incorrectly?
Thank you very much!
In my view time complexity isn't the right way to think about drawing the diagram. It can be done in seconds. What takes time is understanding the data and systems the diagram represents, discussing alternative diagrams with stakeholders, and adapting to the limitations of the available or obtainable data (e.g. some variables cannot be observed). This part takes time.
Given the above, multiple causal diagrams can exist, and may be equally valid if the underlying system is not fully understood.
If the causal diagram is incorrect, any statistical analysis based on it might be invalidated.
However, not drawing a causal diagram doesn't help you. Your statistical analyses may be just as invalid, but without the diagram it's difficult to know either way: Your assumptions are not explicit! This is why I would advocate for always drawing causal diagrams when exploring causal questions with observational data (and in other experiment designs).
We have written an article about this here: https://causalwizard.app/inference/article/causal-diagram
This issue is stale because it has been open for 14 days with no activity.
This issue was closed because it has been inactive for 7 days since being marked as stale.