Simpson's Paradox

A good and accessible article about Simpson's paradox, that's also freely available online is Wang, Wu, Kwan, Tu, and Feng, 2018, "Simpson's Paradox: Examples," from the Shanghai Archives of Psychiatry.

I posted a video about this on my YouTube page. I don't think we have the data to be sure about this yet, but I see all the signs that the common conclusion that SARS-CoV-2 is mostly a threat to the old as a case of Simpson's Paradox.

In short, Simpson's Paradox means that you can get unusual results if you omit variables from your analysis. This is a very easy thing to do in social sciences. In the case of COVID-19, I think a lot of people see the preliminary data that we have and conclude that it's mostly a problem for the old, because mostly old people are dying. But what we are omitting is what, in epidemiology are called confounders, but which in regression analysis in economics we might just call omitted variables. What I suspect is that SARS-CoV-2 deaths during COVID-19 are more about those confounders and less about age than we think.

This means we may see a lot deaths amongst young people who have those confounders. And you all probably know some people at SUU or in the SoB that have some of them.

In addition, young people have a set of confounders that. old people don't have as much, but which can be likely to require ICU care. Specifically, SARS-CoV-2 is a threat to healthy young people who: get in car accidents, attempt suicide, are involved in violent altercations, have problem pregnancies, or find out for the first time that they're severely allergic to something.

******************************************************

And, as long as I have your attention, let me give you an example from macroeconomics and public policy. If you want to go read more on this, check out the article entitled "Surprise  Florida and Texas Excel in Math and Reading Scores" from the New York Times. My copy is corrupted, so I don't have an exact date, but I believe it is from October 2015.

Anyway, most people are certain that the schools are worse in places like Florida and Texas than they are in places like California. That might not be so.

What the data shows is that, within the group of white students, average scores are higher in Texas and Florida than in most other states. And, within the group of African-American students, average scores are higher in Texas and Florda than in most other states. But, no matter where you are, average scores for African-American students are lower than for white students. Except that Texas and Florida have more kids from the weaker group and fewer from the stronger group. And when you just look at the overall average across all students, Texas and Florida are towards the bottom of states ... even though they're better than most if you include the confounding variable of race.

Here's a report from the (definitely) Democratic-oriented, but possibly too old school to be progressive-leaning, Urban Institute noting this same thing.