As I’ve mentioned ad nauseum, I’m not a scientist, but rather a layman with a science jones, one I’ve had from before I could read. So, having read a great deal of science, and of philosophy and history, and, while no mathematician, having at least a feel for numbers and being unafraid of them, I have strong opinions about the use and misuse of science.
But – here’s the thought – whenever I try to explain why something is *obviously* nonsense, I run into the whole issue of what’s obvious to me isn’t obvious to very many people. A sports analogy: I can critique somebody’s jump shooting form at a very high level, but know next to nothing about pitching – I’ve done the one, but not the other. A pitcher who, say, fails to properly rotate his hips (or whatever – ignorant, remember) might be totally obvious to anyone who’s ever tried to pitch, but I’d never catch it. But a jump shooter who fails to get properly squared up and lets his elbow wander? *Obvious*!!! And this is true everywhere: my wife can look at a row of crochet and spot the *obvious* problem, and all I see is a bunch of yarn; my daughter can wing a cake recipe, and correct the batter mid-stream – it’s obvious to her what the problem is. For two of a million such examples.
Back to science. Extreme example: when I read through the abstracts of the first few papers widely reported to *prove* the efficacy of masks in cutting the spread of certain airborne viruses, the flaws of the first few studies were glaringly OBVIOUS. To me. So obvious I never got past the abstracts – why bother?
But then, try to explain it in words laymen can understand. The first study was a meta study, a survey over time of the effects of mask use across a number of states and times frames, which revealed, if I recall, a 2.5% decrease in the rate of case growth in a state where masks had been most rigorously legislated versus a state where mask wearing had been less rigorously called for.
This is obviously stupid, right? Right? What sprang to mind unbidden:
- A vast array of factors figure into case growth – you really think a meta-study is going to properly account for them state by state, such that mask-wearing is going to then account for the remaining difference? How? Bloody unlikely.
- What are the underlying dynamics of the spread of the virus? How can we assume they are the same from state to state, or, indeed, from block to block, such that masks can be credited with effectively changing them?
- Data consistency: measuring numbers across time and space, on different groups with different people doing the measuring in a highly stressful situation? Unlikely to be very good, consistent data. Like, not going to happen.
- Compliance: who cares what the rules are – are people doing as they are told? In any crowd, I’ll see a large range of interpretations of ‘wearing a mask’, from the dude with a paper mask not covering his nose, to people with 2 masks and a clear shield, and everything in between, including my pirate t-shirt scrap hanging down to my sternum. How would you even begin to account for such differences?
- Did we take ALL the states into account? If we left some off, why? The temptation to simply pick a few we like and consciously or otherwise skew the results must be prevented.
- 2.5% You got to be kidding me. Well within the range of ‘noise’.
And so on.
Thus, every time I’ve tried to explain, for example, that not just the list of variables, but order in which they are analyzed, will largely determine the outcome of a regression analysis – um, not so obvious, right? Or that the implicit assumption of homogeneity across patently heterogenous populations – like, for example, talking about *a* population fatality rate – is insane? Clear? Or that scale matters – that you have to have a clear baseline against which to measure change, or your numbers are meaningless?
Anyway, I’ve got my work cut out for me.