Freakonomics M.D. and experiment design


Lucas A. Meyer


July 17, 2022

Data science degrees teach students a lot about how to perform an experiment, but very little about how to set an experiment up.

Freakonomics M.D., besides being interesting by itself, is also a masterclass in experiment design.

Every episode discusses at least one paper about healthcare, and frequently a paper that sits in the intersection between healthcare and economics.

A lot of its greatness has to do with the host, Dr. Bapu Jena. Besides being a medical doctor, Dr. Bapu is also an economist.

What I like the most about Freakonomics M.D. is that Dr. Bapu really takes the time to explain the experiment design of the paper being discussed. He explains why it works, what are the assumptions, confounders, limitations, etc. These are things that are usually confusing for students that are early in their research careers, and in my experience, for most professional data scientists. Many experiment designs are new to me. I find myself full of new ideas after every episode I listen to.

Another thing that the podcast taught me is the importance of knowing some big datasets well (the Medicare database appears in a lot of episodes), but also how to combine the data from these datasets with novel data.

For example, in the very first official episode, they explore whether children’s birthdays increase COVID transmissions. By looking at birthdays and insurance records, one can do an event study of whether having a children birthday in the house increases the odds of having a COVID infection the household afterwards (it does). The episode also discusses robustness checks.

In another episode, they explore whether distractions impact treatment outcomes. They take advantage of the existence of a database with surgery outcomes and another one with surgeon’s birthdays. Making the reasonable assumption that people are distracted on the day of their birthday, they find that surgery outcomes are in fact worse on surgeons’ birthdays.

Overall, I highly recommend this podcast. It’s entertaining, and also useful in many subjects: #healthcare, #economics, #econometrics, and experiment design.

That’s a good value for my time.