Subject: Configuring Effect Direction in Causal Analysis with BOSS and SEM Models

[miaomiao-alt]

I am currently working on analyzing causal relationships among variables using the BOSS algorithm and SEM models. However, I have encountered a question that I was hoping you could help clarify.Specifically, I would like to know if it is possible to set the effect values between variables to be exclusively positive or negative within the software. If such a setting is available, could you please advise on the steps to configure this?

[jdramsey]

When you say "effect" do you mean just one edge? You could set that manually...

[miaomiao-alt]

yes，one edge. Can I set it up in this interface？

[jdramsey]

Yeah whichever one you want to change you can set to, way, B ~ U(0, 1). You should be able to click there and edit directly.

…

On Sun, Nov 3, 2024 at 9:52 PM miaomiao-alt ***@***.***> wrote: yes，one edge. Can I set it up in this interface？ image.png (view on web) <https://github.com/user-attachments/assets/bb25293f-c69e-4c25-91f8-641a7ec87c4c> — Reply to this email directly, view it on GitHub <#1827 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACLFSRZFM6RGKVQ4KRPI5KDZ63OOXAVCNFSM6AAAAABRDJ3KBKVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDINJTG4ZTQNBUGI> . You are receiving this because you commented.Message ID: ***@***.***>

[miaomiao-alt]

So in the search for DAG using the BOSS algorithm, can I specify that all the effect values are positive?

[jdramsey]

Not in BOSS; that's a structure search.

[jdramsey]

If it helps, for a LG SEM model, if the effect is not positive when you estimate it, I believe you can just flip the signs of all edge parameters and make it positive.

[miaomiao-alt]

So in this case, if the effect value on the other side is positive, does it become negative after the transition

[jdramsey]

Not sure what you mean. Keep in mind Boss is trying to optimize BIC, not coefficient values.