Want to join our book club?
We’re reading the new Causal Inference Book by Miguel Hernan and James Robins. The book is forthcoming publication by Chapman & Hall/CRC and it is available for FREE right now to download from Harvard. Every few weeks my colleagues and I meet up to discuss a few chapters. We summarize, discuss, reflect, and hotly debate the tradeoffs in approaches. Overall, the text solidifies many of the evolving best practices in health econometrics and pharmacoepidemiology. It will likely become a new standard textbook used in graduate school training programs around the world.
The goal of our book club is to learn. While draped on couches, and scrolling through the book on our MacBooks, we affirm methods used often in our own research, consider trying approaches proposed that might be superior, and imagine the downstream implications for different applications.
Here are three steps you can follow to be part of our Causal Inference Book Club:
We found chunks of three chapters at a time to be the right amount of meat for a lunchtime book club meeting. There is no reason you have to force yourself go in order either.
After chapters 1-3, we jumped straight to the tempting middle section chapters 12-14 about inverse probability weighting, standardization, the parametric g-formula, and structural nested models. Because we could not wait. It was too cool.
Once you start reading, think and talk with your friends. Here are some of my favorite sections and quotes that I shared with my friends during a book club meeting.
“Not surprisingly, observational methods based on the analogy with a conditionally randomized experiment have been traditionally privileged in disciplines in which this analogy is often reasonable (e.g., epidemiology), whereas instrumental variable methods have been traditionally privileged in disciplines in which observational studies cannot often be conceptualized as conditionally randomized experiments given the measured covariates (e.g., economics).”Causal Inference, Ch 3, pg 27
If we start with what is required for causal inference, it will be easier for the scientific community to vet the rigor of new methods for observational studies.
You don’t know what you don’t know
Causation, association, or prediction?
3. Show up
We will have a *LIVE* virtual book club meeting, thanks to hosting by the ISPOR New Professionals Network. More details coming soon. Until then, start reading. Talk about the #CausalInferenceBook with me @InfectiousEcon and others. The author @_MiguelHernan is actively responding to comments from readers.
Our causal inference book club includes several nerd influencers, celebrity scientists, and young investigators, including:
- Mark Bounthavong, Veterans Affairs and Stanford University
- Ernest Law, Pfizer
- Elisabeth Oehrlein, National Health Council
- Sanket Shaw, Stratevi
- Carrie Bennette, Flatiron Health
Real life as a scientist
It is good to have nerdy scientist friends outside of work. I have some solid ones that answer the phone when I call to ask for advice. Last fall, I observed something sad and unfair in a set of electronic health records. I was curious if a national policy made any difference on the outcome. The idea nagged at me, I really wanted to turn it into a Flatiron Hackathon project.
Even with years of training and multiple degrees, I still couldn’t nail down the best way to design an experiment that would test my hypothesis and show this causes that. I had some good ideas in the shower, because most good ideas happen in the shower. I jumped out and used my favorite window equation pen to capture an possible regression model that might work. I sketched a plot of what I thought the results might look like. But I knew it wasn’t right, or at least that there was a cleaner, stronger way to test the hypothesis and get closer to causal inference.
I called Mark. It was 2 am in New York (me) and 11 pm in California (Mark). I described my data and problem. Together on speakerphone, we explored the tradeoffs of logistic vs Poisson, fixed effects vs random effects, time-varying covariates, and the interactions needed. After 2 hours on the phone, my bathroom mirror was covered and I had a plan. I needed to fit a triple diff-in-diff model.
Math can be hard and fun.
If you don’t have a Mark to call when you have a good idea in the shower, you probably need to download this book as a substitute.
This is the second book pick in our series. Learn more about the discussion of our first book club pick here.
Hernán MA, Robins JM (2019). Causal Inference. Boca Raton: Chapman & Hall/CRC, forthcoming.