regression adjusted estimator for propensity_score_weighting

[adg-ci]

Thank you for developing this package.
I am looking for regression adjusted estimator for propensity_score_weighting.
Is that something the package would be supporting in the future?

Thanks.

[amit-sharma]

That will be a very good addition. If I understand correctly, you are referring to the doubly robust estimator.

It may be a while before we get to it. Would you like to contribute and add it to the package?

[adg-ci]

I am using this package for a project at work. I would have to see how the approval process works for such contributions. Not clear at the moment.

[emrekiciman]

Thank you for considering contributing to DoWhy. In case it is useful for your work to consider, here is the section on contributing to DoWhy that is included on our project page, such as the contributor license agreement and FAQ. Contributing<https://github.com/microsoft/dowhy#id25> This project welcomes contributions and suggestions. For a guide to contributing and a list of all contributors, check out CONTRIBUTING.md<https://github.com/microsoft/dowhy/blob/master/CONTRIBUTING.md>. You can also join the DoWhy development channel on Discord: [discord] <https://discord.gg/cSBGb3vsZb> Most contributions require you to agree to a Contributor License Agreement (CLA) declaring that you have the right to, and actually do, grant us the rights to use your contribution. For details, visit https://cla.microsoft.com<https://cla.microsoft.com/>. When you submit a pull request, a CLA-bot will automatically determine whether you need to provide a CLA and decorate the PR appropriately (e.g., label, comment). Simply follow the instructions provided by the bot. You will only need to do this once across all repos using our CLA. This project has adopted the Microsoft Open Source Code of Conduct<https://opensource.microsoft.com/codeofconduct/>. For more information see the Code of Conduct FAQ<https://opensource.microsoft.com/codeofconduct/faq/> or contact ***@***.******@***.***> with any additional questions or comments. From: adg-ci ***@***.***> Sent: Wednesday, April 20, 2022 11:08 AM To: microsoft/dowhy ***@***.***> Cc: Subscribed ***@***.***> Subject: Re: [microsoft/dowhy] regression adjusted estimator for propensity_score_weighting (Issue #407) I am using this package for a project at work. I would have to see how the approval process works for such contributions. Not clear at the moment. - Reply to this email directly, view it on GitHub<https://nam06.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fmicrosoft%2Fdowhy%2Fissues%2F407%23issuecomment-1104260492&data=05%7C01%7Cemrek%40microsoft.com%7C2735a7f40f3b4316a6ce08da22f8b305%7C72f988bf86f141af91ab2d7cd011db47%7C1%7C0%7C637860748801303912%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=Eimmop8WuNxXtEGViQexLrB%2B1%2BKW0KJryqP3BPq3GHM%3D&reserved=0>, or unsubscribe<https://nam06.safelinks.protection.outlook.com/?url=https%3A%2F%2Fgithub.com%2Fnotifications%2Funsubscribe-auth%2FABNUPUCZXKEQ7MJ53BF2MZDVGBBXZANCNFSM5TJNAB6A&data=05%7C01%7Cemrek%40microsoft.com%7C2735a7f40f3b4316a6ce08da22f8b305%7C72f988bf86f141af91ab2d7cd011db47%7C1%7C0%7C637860748801303912%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=KAWl8MHnXLS7Vjd6YHI3NIkzAnmh4ZNPtwtbtUSqFGc%3D&reserved=0>. You are receiving this because you are subscribed to this thread.Message ID: ***@***.******@***.***>>

[rahulbshrestha]

Hi! I was thinking of working on this issue i.e implement the doubly robust estimator in doWhy. Do you have any pointers so I could get started?

[amit-sharma]

Sure. At a high level, the doubly robust estimator is a combination of the regression estimator and the inverse propensity weighting estimator. So, you will need to create a new Estimator class having these models as part of it (e.g., as attributes).

For the exact way in which these two estimators should be combined, you can refer to, https://academic.oup.com/aje/article/173/7/761/103691. Let me know if you have any questions.

[nparent1]

Hello! I saw this ticket was still open, so I've opened a PR which offers an initial implementation of the doubly robust estimator: #1346, to address the ticket. I also double checked in with Rahul here to make sure it was fine for me to proceed here.

Would appreciate a review!