SSI 2018 New Course Spotlight: Introduction to Causal Inference by Nathaniel Woodward

SSI 2018 New Course Spotlight: Introduction to Causal Inference by Nathaniel Woodward

Instructor Biography

Nathaniel Woodward (né Raley) is an instructor and education researcher at UT Austin. He earned his MS in Statistics from UT in 2016 and he is currently a PhD candidate in Educational Psychology. His research explores how to optimize learning in the classroom for retention across the lifespan through applied cognitive science. In his work, he uses causal inference techniques to evaluate the efficacy of large scale educational interventions. He has worked for SDS as a statistical consultant, and this is his fourth SSI, having served as instructor’s assistant since 2015. Before coming to UT, Nathaniel attended Reed College in Portland, OR, majoring in Biology. He got his start teaching through 1-on-1 tutoring and became obsessed: he has now been teaching in some capacity for over seven years. He lives in south Austin with his wife and their three cats.

What is the main goal of this course?

At bare minimum, everyone will gain an appreciation for the theoretical foundations and practical applications of causal inference! But my goal is a bit loftier: I want to sow the seeds that will change the way you interpret and conduct research, and I will give you all the resources you need to blossom in this direction! We will learn about many techniques, and you will gain experience using them on real data to answer real questions, thus providing a template for you to apply in your own research. Also, in a world of fake news, pay-to-publish, and failure-to-replicate, it is important to be a critical consumer of others’ research too: through the lens of causal inference, you will find that this task is made much easier!

What makes you excited about teaching this course?

What could be more exciting than causation? It is at the heart of the whole scientific enterprise! And it goes way back: Ancient philosophers decided that life was best spent seeking knowledge of “first causes” —Raphael famously depicted them in his School of Athens fresco, which bears the inscription Causaurum Cognitio, exhorting us to “know the causes.” Now, in the 21st century, we are much better equipped to grapple with questions of cause and effect than Plato et al. could ever dream of!

I have been excited about causal inference ever since I started graduate school at UT, but it wasn’t until I took Dr. James Pustejovsky’s course EDP 381 that I understood the full power and beauty of these techniques. Life is messy. 

To say that human behavior is multiply determined is the height of understatement! Get a bunch of people interacting with each other over time, and making causal claims about what they’re doing becomes very, very difficult. Unraveling these vast tangles of influence is an enormous challenge, but we can do it (with enough clean, relevant data)! This course is all about how to do so.

What is the difference between statistical inference and causal inference?

Great question! Statistical inference is about using data from a sample to learn about a population. Now in theory, you could have collected data from the entire population; in this case, you have no more need for statistical inference! You have complete information, the full truth! Causal inference makes use of techniques from statistical inference, but asks fundamentally different questions of the form did X cause Y? Questions like these involve counterfactual states of the world—what-if situations that we can never observe—for which it is impossible to collect data! To estimate the average effect of a treatment, we make inferences about what would have happened if there was no treatment? These are the kinds of question that causal inference is used to tackle, and identification of such effects depends on the ability of our data to let us control for confounding factors.

How much background knowledge or experience in this subject is required to be able to follow the course material?

We are going to start slow (e.g., experiments, correlations) and lay a conceptual foundation (the potential outcomes framework), upon which we will build up to more advanced applications: things like propensity score matching and regression discontinuity analyses. Ideally, you will already be able to run a linear regression and interpret the output. Passing familiarity with logistic regression is helpful, but not required. Some exposure to R/RStudio is also helpful; at minimum, you should have it installed on your computer!

What skills and knowledge can participants expect to acquire by the end of the course?

In this course, the potential outcomes framework will be your key to understanding the power of randomized experiments; this insight will unlock several techniques that are useful for different situations in which you don’t have random assignment (which is often the reality). You will learn when and under what conditions it is appropriate to draw causal conclusions from observational studies, and you will master several important techniques for doing so. To boot, you will learn a bit about general research best-practices (e.g., bootstrapping, robust estimators, cross validation) and your regression chops will get a workout because most of these techniques involve models of this sort.

What was your favorite part of volunteering for SSI in the past? what made you interested in being a lecturer for the SSI program and teaching a brand new course, Introduction to Causal Inference?

SSI is so much fun: it’s hard to pick a favorite part! You are surrounded by all kinds of like-minded people who are dedicating a week of their summer to improving their statistical know-how. It is a really unique thing, and I’ve been lucky to be a part of it all! As much as I’m looking forward to teaching this year (I’m a passionate educator [truth-sharer] and researcher [truth-seeker]; what more needs to be said?), I’ve really relished my previous role as a course assistant because I got to help make everything run smoothly behind the scenes; a botched software installation or a spotty internet connection wasn’t going to hold up the show! I’m looking forward to seeing you at SSI 2018!

The Department of Statistics and Data Sciences at The University of Texas at Austin is proud to host the 11th annual UT Summer Statistics Institute (SSI) May 21 - 24, 2018. Registration is now open!