My name is Mike Famulare and I’ve been working as an epidemiological modeler at the Institute for Disease Modeling for a decade now. In that time, I’ve been part of a lot of things—some you might’ve heard of like the Seattle Flu Study and very stressful days in February 2020 or This Podcast Will Kill You (transcript), and many you probably haven’t (Google Scholar).

During this time, I’ve grown a lot of strong opinions, partially formed and not-so-loosely held, about the practice and philosophy of mathematical modeling in epidemiology. Before the COVID-19 pandemic, I was mostly content to daydream and chat about this stuff over lunches. For about 18 months during the pandemic, when I spent a lot of time working voluntarily1 in close coordination with the Washington state Department of Health, I had a very direct outlet for my thoughts about the best2 ways to help people make good decisions amid uncertainty3. Since withdrawing to better address the enormous toll that work took on my mental health and family4, I’m left with a lot of feelings about how to be a good modeler5 that are bigger and messier than the things I normally do at work can contain. So I need a place to work some stuff out. Hence this blog.

Like all blogs, we’ll see what it becomes, if anything. But here’s some of what I expect it to be about, at least at the start.

  • Quantified uncertainty matters, and a good understanding of uncertainty often makes it easier to know what you should do next.

  • Everything that looks like noise is variation that can be modeled, except for stuff you can trace back to thermodynamics6. But do you need to model it? When?

  • Predicting things you haven’t yet measured is the best way to test your understanding. But there are lots of ways to predict lots of things. Choose your objectives wisely.

  • Nothing beats a good measurement, except a great theory7.

  • Some studies, experiments, models, and papers are really, really great and we should celebrate them!

  • Wicked problems

  • We can do this with my model, but should we?

  • Experiments in footnotes, content, writing style, to help me find my voice anew.

If this blog has a role model, it might be this post by John Salvatier, Reality has a surprising amount of detail. Details matter, but you never know which details matter for which questions until you understand the structures that hold them together. Nothing is one thing.

  1. Voluntarily-ish? My day job became dominated by public health work, laundered through “this is best way I can be locked in to COVID-19 dynamics and do generalizable science.” I’m very grateful to have had a whole lot of latitude within my roles at IDM. ↩

  2. LOL. Somehow I hope starting a blog will be a treatment for arrogance? ↩

  3. and, within the constraints of what’s possible in the US, we’ve done better than most (Age-adjusted COVID-19 deaths per capita across US states as of April 27, 2022). ↩

  4. I’m scared to write about this in public, but I hope that changes. ↩

  5. The treatment isn’t working, Doc! ↩

  6. One could spend a lifetime thinking about all the ways this is true and false. ↩

  7. In the physics sense, where a theory is a model that’s passed so many different kinds of quantitative tests that you are more likely to get things right if you trust the theory over any one experiment or study, except where the theory is known to be weak. This is messier in the biosciences, but it’s still a good rule, I promise you. ↩

<
Blog Archive
Archive of all previous blog posts
>
Next Post
The best tools are fit for purpose