# Microsoft transitioning Causal Inference library to Open Source community

**URL:** <https://community.intuitivebayes.com/t/microsoft-transitioning-causal-inference-library-to-open-source-community/166>\
**Category:** Causal Inference Book Club\
**Created:** [June 1, 2022, 11:20pm UTC](https://community.intuitivebayes.com/t/microsoft-transitioning-causal-inference-library-to-open-source-community/166 "2022-06-01T23:20:13Z")\
**Posts on this page:** 4\
**Page:** 1

<div class="post-metadata">

**Author:** ![RavinKumar](https://yyz1.discourse-cdn.com/flex007/user_avatar/community.intuitivebayes.com/ravinkumar/32/13_2.png) [@RavinKumar](https://community.intuitivebayes.com/u/RavinKumar)\
**Post date:** [June 1, 2022, 11:20pm UTC](https://community.intuitivebayes.com/t/microsoft-transitioning-causal-inference-library-to-open-source-community/166/1 "2022-06-01T23:20:13Z")

</div>

Copying a tweet from @marksimi here

Looks like Microsoft Research is also excited about Causal Inference.

Thanks for sharing @marksimi

> <https://twitter.com/marksimi/status/1532073428039385088>
>
> Microsoft Research @MSFTResearch

---

<div class="post-metadata">

**Author:** ![mancunian1792](https://avatars.discourse-cdn.com/v4/letter/m/e79b87/32.png) [@mancunian1792](https://community.intuitivebayes.com/u/mancunian1792)\
**Post date:** [June 2, 2022, 3:07pm UTC](https://community.intuitivebayes.com/t/microsoft-transitioning-causal-inference-library-to-open-source-community/166/2 "2022-06-02T15:07:16Z")

</div>

If you haven’t checked out Microsoft’s [End to end causal inference system](https://www.microsoft.com/en-us/research/publication/deep-end-to-end-causal-inference/), this is a layer on top of DoWhy. I recently attended a workshop where they showcased the tool in its current form. It makes causal discovery more intuitive and tries to help people go from some knowledge in CI to use the tool and evaluate causal queries

---

<div class="post-metadata">

**Author:** ![marcin.elantkowski](https://yyz1.discourse-cdn.com/flex007/user_avatar/community.intuitivebayes.com/marcin.elantkowski/32/71_2.png) [@marcin.elantkowski](https://community.intuitivebayes.com/u/marcin.elantkowski)\
**Post date:** [June 8, 2022, 12:57pm UTC](https://community.intuitivebayes.com/t/microsoft-transitioning-causal-inference-library-to-open-source-community/166/3 "2022-06-08T12:57:44Z")

</div>

Do you know what is the relationship between `DoWhy` and PPLs like `PyMC`?

In a PPL we can define a DAG, run inference, and evaluate counterfactuals / apply do-operators.  
Is doing this in `PyMC` any different than doing it in `DoWhy`?

I can see that `DoWhy` have some extra features (like discovering causal structure) but except for that, what are the cases in which I would want to define a DAG in `DoWhy`, instead of doing so in `PyMC` / `numpyro`?

---

<div class="post-metadata">

**Author:** ![mancunian1792](https://avatars.discourse-cdn.com/v4/letter/m/e79b87/32.png) [@mancunian1792](https://community.intuitivebayes.com/u/mancunian1792)\
**Post date:** [June 8, 2022, 5:19pm UTC](https://community.intuitivebayes.com/t/microsoft-transitioning-causal-inference-library-to-open-source-community/166/4 "2022-06-08T17:19:00Z")

</div>

As far as I understand, doWhy has API’s to do both model ( assuming structure on data generating process using DAG’s) based causal inference and causal inference on observational data ( methods to incorporate potential outcomes framework ). I believe this book covers mostly on the potential outcomes framework and other methods to do CI on observational data. And as you mentioned it has API’s to do causal discovery and identification of statistical estimands for the causal query in question!

I have not used PYMC to do CI with DAGS but have used pyro! I can share few notebooks that explains how to do the above with Pyro, if anyone is interested. Would be interesting to compare across tools and see which one is more intuitive.