Episode 31 - The Problem of Causality with @ShirinMojarad

We do things every in the hope that it will cause something good to happen as a result. And we're also bombarded with information trying to convince us on what to buy, who to vote for, and how to live. But most of the time, it's very difficult to tell whether A causes B, or if they are merely correlated. I talk to subject matter expert Shirin Mojarad of McGraw-Hill to break down how scientists can detect causality and what to look for in a causality study.

Links

First, if you want to learn more about Shirin’s recent work at McGraw Hill and the ALEKS Study, read Shirin’s article on Efficacy in Education and ALEKS and a recent BI article about the effort.

For the more technical treatment, read Shirin’s paper. It’s only 8 pages, but packed with information on how these types of studies are performed.

Check out Andrew Ng’s Slides on what data scientists should know about Deep Learning. Slide 30 contains the chart Shirin mentioned with a point where algorithm research has higher marginal value than data acquisition.

The Book of Why by Judea Pearl – recommended for a general audience. Go to his website for a free read of the first couple chapters.

For a more technical text, check out Andrew Gelman’s Book with causal models in chapters 9 and 10.

Maryam Mirzakhani, winner of the fields medal in mathematics, wikipedia bio and New York Times Obit.

Here’s an article about how pirates and Global Warming correlate, and here’s a whole big list of fun correlations sent in by a Local Maximum listener.

About Shirin Mojarad

Find Shirin on LinkedIn and Twitter

Full bio:

Dr. Shirin Mojarad is currently a Lead Data Scientist at McGraw-Hill Education. She was formerly a senior analytics specialist in the Advanced Analytics team at Canadian Imperial Bank of Commerce (CIBC) and a data mining consultant with a leading software company in predictive analytics.

She is an expert in navigating and deriving insight from large datasets and analyze behavioral patterns using advanced statistical modeling and data mining techniques. Shirin’s expertise lies in drawing causal inferences in observational studies and applying machine learning to the field of causal inference, which she has applied at McGraw-Hill Education and published in international conferences.

Shirin received her Ph.D. in Electrical Engineering and her M.Sc. in Communications and Signal Processing from Newcastle University U.K., where she specialized in predictive modeling and artificial neural networks. Shirin is the founder and leader of three Boston based Meetup groups including Women in Big Data, So You Want To Be A Data Scientist, and Learning Analytics and Educational Data Mining. She has led several workshops on Python, causal inference, and data mining in international conferences including Learning Analytics and Knowledge Conference, European Conference on Technology Enhanced Learning, Chief Data and Analytics Officer, and Big Data Summit. She has been invited as the subject matter expert on several data and analytics panels including Harvard Big Data panel, Chief Analytics and Data Officer analytics roadmap and implementation panels.

Previous Episodes

Episode 27 about the idea of Big Data vs Big Algorithm.
Episode 21 on how to figure out what to believe.
Episode 17 and Episode 6 on advertising effects on elections.