@soboleffspaces I have a question: given a DAG that is compatible with an observational dataset. But I have an extra variable, that is not yet part of this DAG, for which I do not have the data yet. What can be said about the stability of this DAG if later on this var is added?
1) DAG represents knowledge, rather than data. 2) The variable's inclusion in the graph is often determined by the level of abstraction. 3) Our assumption is that the dependence of pa(X) -> X remains unchanged when we modify the graph.
Does this mean that the existing edges would not change due to the inclusion of this new variable? Only new edges could be add, but the old ones would not change?
Well, DAG may evolve as we learn more. But adding a node shouldnโt affect the existing arrows, unless of course you aggregate some nodes๐๐ป
Not me. Itโs the graph by David Eppstein commons.m.wikimedia.org/wiki/User:Daviโฆ
This is exactly what I mean! So I expect if I learn the DAG from obs data without the extra var and then learn a new DAGโ on the data with the extra var, that DAGโ has exactly all the edges in original DAG, perhaps with extra edges in DAGโ coming from the extra var. Do you agree?
#DependencyGraph represents knowledge, not data. We use #CausalDiagram (dependency graph with designated cause and effect nodes) to identify non-causal paths and classify them as either open or blocked. This leads an adjustment set. #causaltwitter sobolevspace.com/teaching/f/shoโฆ
United States ํธ๋ ๋
- 1. #WWERaw 71.2K posts
- 2. Dolph Ziggler 8,526 posts
- 3. Cowboys 33.1K posts
- 4. Raiders 50.8K posts
- 5. Sheamus 6,488 posts
- 6. CeeDee 5,966 posts
- 7. John Cena 40.7K posts
- 8. #RawOnNetflix 1,842 posts
- 9. Geno 8,058 posts
- 10. Quinnen Williams 1,908 posts
- 11. #RaiderNation 2,602 posts
- 12. Rey Mysterio 4,825 posts
- 13. Maxx Crosby 1,093 posts
- 14. Nikki Bella 3,607 posts
- 15. #MondayNightFootball N/A
- 16. Schotty N/A
- 17. Jeanty 2,397 posts
- 18. Survivor Series 5,603 posts
- 19. Monday Night Raw 7,873 posts
- 20. Cam Skattebo 1,944 posts
Something went wrong.
Something went wrong.