Timeline for Causal modeling and DAGs in Python - where to start and what are the best sources?
Current License: CC BY-SA 4.0
10 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Oct 27, 2021 at 18:00 | history | tweeted | twitter.com/StackStats/status/1453421269324636162 | ||
| Oct 27, 2021 at 16:01 | comment | added | Adrian Keister | Well, as I said, that link above seems good - it's at a suitable, intro level. I think I might well avail myself of it! Modeling is a highly iterative process: you try something and see how well it works. xkcd.com/1838 I will say this: it's much easier to see if a given model is off by a lot, than to show that it works really well. You can just start with linear regression: you control for variables simply by including them on the RHS. But, as Pearl has pointed out, be careful what you control for. Avoid controlling for mediators! | |
| Oct 27, 2021 at 15:57 | comment | added | Anakin Skywalker | @AdrianKeister, thanks, I am re-reading The Book of Why right now. It is a mandatory reading on Tatooine. Very intuitive. I also tried Primer. Have you seen any full stack tutorials how to go from a DAG to a model and its evaluation, ideally in Python. I have not. Only some disaggregated pieces. | |
| Oct 27, 2021 at 15:54 | comment | added | Adrian Keister | Yes, the DAG is a visualization, and it goes hand-in-hand with a SCM, although there are MANY things you can deduce about a system merely from its DAG. I also found Pearl's Causality book extremely difficult - I've only been able to force my way through a small part of it. But he has easier books that might help. I recommend reading these, in order: The Book of Why, and Causal Inference in Statistics: A Primer. That's a great link you have. Definitely might help as well. | |
| Oct 27, 2021 at 15:50 | comment | added | Anakin Skywalker | @AdrianKeister, many thanks. I have this book as well his other books. But they are tough as laser sword of Darth Vader. Do I understand correctly that DAG is a visualization and I need to build a model behind, controlling for some variables. Do you have any good tips about resources? I can imagine I can build a regression model, but not sure because I do not see some intuitive tutorials. This might help, no? theeffectbook.net/the-toolbox.html | |
| Oct 27, 2021 at 15:44 | comment | added | Adrian Keister | Ah, I see. We'd better work quickly before you turn to the Dark Side. When you get to the specifics of DAGs, you simply ask yourself, "Does this variable affect that one?" of every variable-variable pair. That's usually the easier part, though it can be highly non-trivial. Pearl's book Causality: Models, Reasoning, and Inference has an algorithm that can help discover DAGs based on real-world data. Getting the Structural Causal Models is harder, though. Your best bet is domain experts: ask them and see what they come up with. | |
| Oct 27, 2021 at 15:37 | comment | added | Anakin Skywalker | @AdrianKeister, I was assigned to a Death Star causality project and need to build a real model, using DAG, showing the causal effect between money spent and events on the number of activities. If you can help to fight against the evil, I would greatly appreciate it :) | |
| Oct 27, 2021 at 15:34 | comment | added | Adrian Keister | Are you just trying to get a model to play around with? Or are you trying to model a real-world scenario? I'd imagine living on Tatooine, it might be difficult to relate to those of us here on earth. ;-) | |
| Oct 27, 2021 at 12:48 | history | edited | Anakin Skywalker | CC BY-SA 4.0 |
added 29 characters in body
|
| Oct 27, 2021 at 12:32 | history | asked | Anakin Skywalker | CC BY-SA 4.0 |