An award for extreme geekiness

Transparent trophies are surprisingly difficult to photograph with the text legibly focused. Copyright © 2019 Wayne and Elizabeth Saewyc.

Elizabeth really and truly did win an award for doing something outrageously intellectual/academic. It is titled the “2019 Robert H. DuRant Award for Statistical Rigor and Innovation in Adolescent Health Research”, which is enough to make one run and hide.

Notwithstanding, the title of the paper receiving the award is even worse, and therefore nearly deserves to get such an award.

In very simple English, Elizabeth and a crew of stats geeks spent about 5 years working out and testing a way to use existing datasets to prove, with about the same rigor as a Randomized Clinical Trial (RCT), ‘causal inference.’

If you have ever enjoyed some Tyler Vigen’s Spurious Correlations then you already have some clue as to why the RCT is the gold standard showing that Y is because of X. Well, Elizabeth’s team came up with something a lot cheaper, and is about as good.

Her particular need was to show that a school based intervention did, or did not, do what it was expected to do. And to measure just how much the effect of the intervention is. Anyone remember the D.A.R.E. program in the USA? Anyone know if it really works to keep youth away from drugs and violence? Well, no, because no one could afford to test it widely, before and after, so the country has spent gazillions on a program which we do not know if it works.

Now we could find out.

The method uses already existing data sets, like those semi-regular surveys of students which have been conducted pretty much since the beginning of formal education in the USA. Through a painfully complex process the SLEPHI method (of course there has to be a completely not-catchy acronym or it would not be geeky enough!) balances out schools and time frames to get comparable data of groups who did, and did not, receive the intervention being studied, and how they changed from before, during, and after the intervention.

For a little while I felt transported back in time because this method took 5 hours to run on each question, just like running normal stats did back in the 80s and 90s.

Anyway. I am partnered with an extreme geek, who has the trophy-ish thing to prove it.

NB: This was actually drafted the day  the award got home, but for some reason was lost in Draft purgatory for a while.