class: center, middle, inverse, title-slide .title[ # Double ML: Causal Inference based on ML ] .subtitle[ ## Part IV: Outlook, Discussion and Conclusion ] .author[ ### uai2022, August 1, 2022, Eindhoven ] .date[ ### Philipp Bach, Martin Spindler (UHH & Economic AI) Collaborators: Victor Chernozhukov (MIT), Malte Kurz (TUM) ] --- class: inverse center middle, hide-logo # Recap --- ## Recap <br> .center[ #### Congratulations! You made your first steps in causal machine learning with `DoubleML` ] -- <br> * Today we have covered + the challenges of causal machine learning + the basics of double machine learning + an introduction to the implementation with `DoubleML` + hands-on examples (price elasticity estimation, AB test) --- class: inverse center middle, hide-logo # Outlook <center><img src="figures/DoubleML_Rhino_1000x1000.png" alt="The logo of the DoubleML package - a double-headed rhino" height="250px" /></center> --- ## Outlook: What's next? #### What's next? * Continue your learning journey and visit our [user guide](https://docs.doubleml.org/stable/guide/guide.html) to learn more about double machine learning, for example, + sample splitting, + hyperparameter tuning, + simultaneous inference, + clustering standard errors, + other model classes and score functions * In case you find bugs or want to start or contribute to a discussion, visit our GitHub repository + **https://github.com/DoubleML/doubleml-for-py/issues** + **https://github.com/DoubleML/doubleml-for-py/discussions** --- ## Extending DoubleML We are currently working or planning various extensions of the `DoubleML` package .pull-left[ #### Double machine learning * Extensions to heterogeneous treatment effects, CATEs and GATEs * Extension of built-in resampling schemes * New model classes and extensions of current classes + Difference-in-Difference estimators + Categorical treatment + Partially linear IV models + AutoDML * New examples in our [**gallery**](https://docs.doubleml.org/stable/examples/index.html) ] .pull-right[ #### Python * Integration of more learners (`pytorch`, `keras`, ...) * Model diagnostics based on `scikit-learn` tools * ... ] --- ## Contributing to DoubleML * We welcome contributions to `DoubleML` + Adding model classes, based on our [**model template**](https://github.com/DoubleML/doubleml-docs/blob/master/model_templates/double_ml_model_template.py) + Change to model specific components, like nuisance estimation, scores, ... + Add your replicable example to our [**gallery**](https://docs.doubleml.org/stable/examples/index.html) + Contribute to [**discussions**](https://github.com/DoubleML/doubleml-for-py/discussions) * Report bugs <br> `\(\rightarrow\)` [**Contributing Guidelines**](https://github.com/DoubleML/doubleml-for-py/blob/master/CONTRIBUTING.md) with additional information .center[ <br> **https://github.com/DoubleML/doubleml-for-py/blob/master/CONTRIBUTING.md** ] --- ## Thank you, uai2022! We appreciate your feedback .center[ `\(\rightarrow\)` **https://forms.gle/otQThHgq6nYHrYzeA** <center><img src="figures/qr.png" alt="A qr code to the feedback form" height="200px" /></center> ] <br> .pull-left[ As a little "**thank you**", we will send you the official `DoubleML` hexagon sticker ] .pull-right[ <center><img src="figures/DoubleML_Rhino_1000x1000.png" alt="The logo of the DoubleML package - a double-headed rhino" height="150px" /></center> ] --- ## Thank you, uai2022! In case you have comments or questions, feel free to contact us <br> .center[ philipp.bach@uni-hamburg.de martin.spindler@uni-hamburg.de malte.kurz@uni-hamburg.de ] <br> For knowledge transfer, industry collaborations, and trainings, check out and feel free to reach out .center[ www.economicai.com spindler@economicai.com <center><img src="figures/economic_ai.png" alt="The logo of the EconomicAI package - a double-headed rhino" height="50px" /></center> ] --- ## Thank you uai2022 ### Acknowledgements * We'd like to thank **Malte Kurz** and **Victor Chernozhukov** for their great collaboration, Sven Klaassen, Jannis Kück, Carles Riera, Oliver Schacht for contributing and collaborating and student helpers for support during the preparation of the tutorial. * We'd like to thank the organisers of the [**2019 ACIC Data Challenge**](https://sites.google.com/view/acic2019datachallenge/data-challenge) for sharing the DGPs and granting us the right to use and distribute data sets. --- class: inverse center middle, hide-logo # References --- ## References #### Double Machine Learning Approach * Chernozhukov, V., Chetverikov, D., Demirer, M., Duflo, E., Hansen, C., Newey, W. and Robins, J. (2018), Double/debiased machine learning for treatment and structural parameters. The Econometrics Journal, 21: C1-C68, doi:10.1111/ectj.12097. * Chernozhukov, V., Hansen, C., Kallus, N., Spindler, M., and Syrgkanis, V. (forthcoming), Applied Causal Inference Powered by ML and AI. #### DoubleML Package for Python and R * Bach, P., Chernozhukov, V., Kurz, M. S., and Spindler, M. (2021), DoubleML - An Object-Oriented Implementation of Double Machine Learning in R, [arXiv:2103.09603](https://arxiv.org/abs/2103.09603). * Bach, P., Chernozhukov, V., Kurz, M. S., and Spindler, M. (2022), DoubleML - An Object-Oriented Implementation of Double Machine Learning in Python, Journal of Machine Learning Research, 23(53): 1-6, https://www.jmlr.org/papers/v23/21-0862.html.