In this blog post, I want to address a subject that I have been avoiding blogging about for a real long time. While I'm by no means a typical low-carb proponent, as an engineer I feel strongly there is no one size fits all diet fit for everyone, or even, as for me fit for all seasons. As I've written about many times before, I'm very much a data guy. Computational statistics, causal inference, probability and uncertainty theory, and next to that, my own control-feedback theory driven N=1 experiments, are the basis for my perspective on nutrition. And looking at the first of those, there are serious issues with the field of nutrition.
Nutrition, as a field of science, is plagued by poorly designed studies, horribly flawed statistical analysis, causal claims and data adjustment methods that ignore decades of advancement in causal inference, political, commercial, and animal right agendas, and I could go on. So when someone comes along, an outsider, looks at the available data, someone schooled in data science, modern causal inference, or just simply someone with a solid background in mathematics or engineering, that person can potentially jolt loose a part of the field stuck between data, politics and narrative based science, and push it back into the direction where most of the rest of science has moved way beyond the horizon already.
From that perspective, over the years, I've ended up following quite a few, as I thought then, bright lights, in the Low Carb community. As I thought. Because, as it turns out, a few of these bright lights turned out to be nothing more than blind squirrels that just somehow managed to stumble into a nut; Once. What happened for me to find out these bright lights were in fact blind squirrels? These low-carb folks with their amazing insights into health and nutrition started blogging, vlogging and tweeting about COVID-19.
My first response was to gently try to guide them back to mathematical rigor, when they started making unsubstantiated claims based on limited data availability at that time. Showing them things like how small differences in lag between demographic groups could skew early demographics. They didn't listen, and yes, the demographics shifted.
Eventually some new self-proclaimed virology experts grew tired of their new field and moved back to nutrition, or made COVID a side issue they rarely touched anymore. But some didn't. One in particularly started seeking backup from a Nobel Prize winning scientist, who helped him feel confident, and from that point on, this blind squirrel that had so far been going by touch, started acting like he wasn't blind at all, frantically running from place to place, bumping into stuff, falling down, getting up and acting as if nothing happened.
Still, not wanting to make the blind squirrel look bad, after all, he had done amazing work on nutrition and health, and even if his insights on these were just the first and possibly last nut our blind squirrel had discovered, it was still a nut deserving of credibility, even if the squirrel that found it was not, I continued trying to give little friendly pokes, hoping the squirrel wasn't completely blind after all.
Didn't help. The blind squirrel went in overdrive. Claims getting crazier and crazier. And with his standing in the community, and with the credibility his Nobel-Prize winning scientist gave him and a few like him, more and more of the low-carb community started jumping on the crazy train.
So what kind of claims are the blind low-carb squirrels making? And how do we know they are wrong?
claim one: The pandemic is over
This one is easy enough to debunk. Here is the week-averaged daily mortality curve for the whole planet:
claim two: There is no secondary wave in Europe
Again, looking at the mortality data on a log-y scale, just one look makes this a ridiculous claim.
We've seen two full doublings in the last 10 weeks, not as many as we saw in February and march, but most definitely a secondary wave. And we are only three doublings removed right now from reaching the daily body count we had at the peak.
It's interesting to see how the blind squirrel are trying to keep this part of the narrative alive. They started out claiming the rise in cases was due to the increased testing and the intrinsic number of false positives the tests yield. There was no increase in deaths. Even if the log-y graph already showed a distinct rise, they insisted the use of a lin-y graph. And for inexplicable reasons, the LC masses just gobbled up the bullshit narrative and went: "OK".
But then even the lin-y scale started showing the secondary wave was real and not just what they refer to with the made up term casedemic, then all of a sudden we shouldn't look at the lin-y graphs of mortality anymore. No we should be looking at excess deaths now.
claim three: Europe is close to herd immunity
There are two easy ways to look at this. The Netherlands so far has been one of the hardest hit countries in Europe in terms of deaths per million. Lets look at the map of the Netherlands and color in the deaths per million per municipality.
Notice how there is almost an order of magnitude difference in the number of deaths per million in the hardest hit muncipalities of this hard hit little country, and important densely populated muncipalities in the Dutch Randstad area?
But we can also look at it from another side. Back to the European mortality graph. Let's compare slopes.
Lets start with the steepest 5 days of the secondary wave:
Now lets compare that to the point on the primary wave where we were seeing the same numbers.
We see that Rt at a time of minimal measures was around 2.3, and recently despite of measures and all countries being on edge, this virus still has the capacity for at least jumping back up to an Rt of 1.6. This means that if this currenly is the peek of what this virus is capable of, the amount of non-immune people in the population nor compared to march is at least 70%. By these figures the number needed for herd immunity would be about 30%, so best case scenario, we are three seventh of the deaths needed for herd immunity. But as the Randstad figures show, these numbers seem quite optimistic.
Claim four: All curves are basically the same
This claim, for anyone who has ever used statistical software to curve fit data, should ring suspect. Unfortunately it's a claim made by a Nobel Prize winning scientist, and that fact makes many people, including myself, think: Am I missing something?
I'm an engineer, not a mathematician. A data engineer with a decent grasp of math. An engineer who uses data fitting to curves on a daily basis for the better part of two decades, but still not a mathematician. So obviously me and everyone who has ever fitted data to curves is wrong about the whole concept of curve fitting and the Nobel Prize winner must be right? Unfortunately for the venerable scientists, we aren't missing something, he is. Fitting COVID data to a so called Gompertz curve isn't a particularly wrong idea. The first derivative of the Gompertz function is a function where the exponent of growth changes (decreases) over time, as time progresses. The function then moves from exponential growth to exponential shrinkage and doesn't come back to one of growth. This doesn't need to be a problem, because composite fitting allows you to fit with multiple Gompertz curves. Fit with one and most curves are under fitted. Fit with a hundred and your curve is overfitted. Just like fitting with polynomial curves, Gaussian curves, or exponential functions using a polynomial as exponent, curve fitting doesn't explain or predicts, it just fits. Somehow though our venerable Nobel Prize winner and our blind squirrel have both gotten it into their head that an under fitted Gompertz curve somehow predicts or explains anything. Not only does it not, single Gompertz derivatives aren't even a better fit than the alternatives.
Claim five: The only way out is through herd immunity
This one is simple too: China
Ignore the box shapes thingy. Its just a correction spread out over 7 days by our moving weekly averaging process.
China crushed their curve. They crushed their curve quick and hard. How? Not sure. What did China know or do that we didn't? Many things, probably. One possibility, they acted based on the knowledge that the virus is airborne. They told us in February it was. Europe and North America ignored this for some reason and only now are slowly coming around. The CDC in the Netherlands is still officially rejecting it. Germany isn't. Could be the reason why China crushed it. Main point though, China showed us that the virus can be crushed without the need to reach herd immunity.
Conclusions
Hope the above showed with five (of many) claims made by our Low Carb blind squirrels, that it really is not that hard to poke holes in the narrative of the data and math blind. While it may be surprising these squirrels were able to find that one amazing nut in the past, sometimes it is just big numbers and survival bias that has us take people serious for a one time moment of brilliance that has long passed. Be aware of blind squirrels no matter how big and shiny the nut they found looks. I chose to neither disclose the names of the people involved or hide who I am talking about in this blog post. I think many of you will know exactly who the primary blind squirrel is I'm talking about. But it really doesn't matter. There are many more like him. One hit wonders who gained some credability with a nut they found and now spread their ignorance with false confidence and credability.
Be aware of blind squirrels, especialy when COVID19 is involved.