Python and the Holy Grail of Causal Inference - Dennis Ramondt, Huib Keemink

PyData Amsterdam 2018
Causal Inference, AKA how effective is your new product, policy or feature? Inspired by A\B testing in tech, organizations have turned to randomized testing. However, randomization often fails, leaving us in a biased reality. Join us on our quest to dispel myths about randomized testing and build practical models for effect measurement in business situations, in this Eneco-Heineken joint talk.
Slides: www.slideshare.net/PyData/pyt...
--
www.pydata.org
PyData is an educational program of NumFOCUS, a 501(c)3 non-profit organization in the United States. PyData provides a forum for the international community of users and developers of data analysis tools to share ideas and learn from each other. The global PyData network promotes discussion of best practices, new approaches, and emerging technologies for data management, processing, analytics, and visualization. PyData communities approach data science using many languages, including (but not limited to) Python, Julia, and R.
PyData conferences aim to be accessible and community-driven, with novice to advanced level presentations. PyData tutorials and talks bring attendees the latest project features along with cutting-edge use cases.
0:00 Introduction
0:36 What is Causal Inference
1:18 "Slogan" for the Talk
1:35 Typical Randomized Testing
2:15 Use Case 1: Heat Pump Savings
2:45 Measurement Data: Daily Gas Usage ~ Outside Temp
3:32 Experiment Setup: Heat Pump Savings
5:40 Fixing Group Imbalance
6:50 Propensity Score Matching
7:47 Recap of Heat Pump Use Case (Correcting Assumptions and Limitations)
9:07 Use Case 2: Effect of Cooler Placement
9:40 Experiment Setup: Cooler Placement
11:47 Problem 1: Test and Control Groups are Statistically Different
13:17 Code Example: Single Covariate Simulated Example
14:01 Treatment Effect Correction
16:14 Code Example: Treatment Effect Correction
16:28 Assumptions: Conditional Mean Independence & Conditional Independence
18:44 Tree Ensemble: Virtual Twins
20:21 Results: Cooler Placement
22:18 Talk Recap
24:50 Q&A 1
26:38 Q&A 2
27:32 Q&A 3
29:00 Q&A 4
30:29 Q&A 5
S/o to github.com/trfore for the video timestamps!
Want to help add timestamps to our KZbin videos to help with discoverability? Find out more here: github.com/numfocus/KZbinVi...

[@fc1984fc]

6:49 could one not use clustering to find the closest points in the multidimensional space and use them as partial matches? And to increase the chances of getting a higher number of neighbours, one could also discretise the values in the raw data by binning.