www.pydata.org
We learn about the world from data, drawing on a broad array of statistical and inferential tools. The problem is that causal reasoning is needed to answer many of our questions, but few data scientists have this in their skill set. This talk will give a high-level introduction to aspects of causal reasoning and how it is complemented by Bayesian inference. A worked example will be given of how to answer what-if questions.
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.
00:00 Introduction to talk
00:31 Package announcement
00:33 Speaker introduction
01:13 Causal inference is trending
01:22 Google Trends on Causal inference
01:51 Is it convincing enough?
02:41 Bayesian model on the trends using PyMC
03:45 Hype Cycle for emerging tech by Gartner
04:10 Difference between Statistical relationships and Causal relationships
06:55 Observational study on causal relationship between Tea and Death
09:00 Confounding variables in our study
09:33 Randomized control trial (RCT)
10:25 Can you model confounding variables and not randomize?
12:17 Randomization is very effective
12:35 Randomized control trials can be problematic
14:56 Quasi-Experimentation by Charles S. Reichardt
16:08 CausalPy package
16:44 What does CausalPy do?
16:50 Example: What was the causal impact of Brexit?
19:00 Normalized GDP
19:55 What do we not have on this graph?
21:18 Fitting the model
22:32 Synthetic control method in CausalPy
23:28 Visualizing the output
26:03 Other features of CausalPy
26:10 Interrupted time series
26:52 Regression discontinuity
27:49 Difference in differences
28:04 Did my advertising budget cause more sales?
29:41 Summary
30:54 Q/A Any suggested resource to properly design RCT?
31:56 Q/A Why didn't you use a diff-diff model?
32:51 Q/A Training a ML model to predict pre-treatment GDP of UK
34:07 Q/A How is CausalPy related to CausalImpact
35:07 Q/A Interrupted time series and regression discontinuity