They bypass skin color by just measuring levels of Vitamin D directly. This study doesn't care how you got the Vitamin D, it just cares how much you have.
There are tons of problems with this study, but failure to look at skin color isn't one of them.
People with darker skin will, ceteris paribus, have lower vitamin D levels. And all else is not equal.
Even if skin color is a predictor of vitamin D levels, this study doesn't need to predict vitamin D levels because they used measured vitamin D levels.
If darker-skinned people have, on average, lower vitamin D levels, then measuring vitamin D levels proxies for skin color.
So if the underlying cause is e.g. access to healthcare or stress or exposure to airborne particulates, and those correlate with skin color, those will also correlate with vitamin D levels.
One could thus find vitamin D levels relate strongly to outcomes. But increasing vitamin D has zero effect on the outcome. Because it's not the cause. (We see this a lot with poverty correlates in education.)
"Proxies for" doesn't have a scientific meaning that I'm aware of, so I can't be sure that I understand what you're saying.
If what you're saying is that vitamin D levels are correlated with skin color, true, but irrelevant to this study.
If what you're saying is that the (poorly substantiated) correlation between Covid19 and vitamin D can be used to claim a correlation between Covid19 and skin color: no, because it's quite possible that vitamin D levels have no effect on Covid19 in people with darker skin, but a huge effect on Covid19 in people with light skin. You also have to keep in mind that your errors multiply across different correlations: the error margin of correlating Covid 19 with skin color via this method is the error margin of correlating Covid19 with vitamin D and the error margi of correlating vitamin D with skin color.
If what you're saying is that the (poorly substantiated) correlation between Covid19 and vitamin D can be used to establish a causal relation between Covid 19 and skin color, no, because in addition to the previous problems, you're ignoring that there at least 10 confounding factors besides skin color which also have correlation with vitamin D levels:
1. Latitude
2. How much time you spend outside
3. Which time you spend outside
4. Clothing
5. Sunscreen
6. Liver function
7. Diet
8. Supplementation
9. Cloud cover over time
10. Season
If what you're saying is that this study can be used to establish a causal relation between Covid19 and skin color: in addition to all the previous problems, this study can't even really be used to establish a correlation between vitamin D and Covid19, so do we really need to pile on more logical fallacies than are already present?
It's a term, roughly, for intermediate variables [1]. Vitamin D levels are a likely proxy for a host of race-variant socioeconomic factors.
> In statistics, a proxy or proxy variable is a variable that is not in itself directly relevant, but that serves in place of an unobservable or immeasurable variable. In order for a variable to be a good proxy, it must have a close correlation, not necessarily linear, with the variable of interest. This correlation might be either positive or negative.
> Proxy variable must relate to unobserved variable, must correlate with disturbance, and must not correlate with regressors once disturbance is controlled for.
Skin color isn't an unobservable or immeasurable variable, and it does correlate with regressors once disturbance is controlled for.
To account for this, you would need to control for skin color.
They are doing this by using public measures of country-wide serum-based Vitamin D deficiency rates and COVID-19 outcomes.
How is "where the Vitamin D comes from" relevant at all to this particular correlation? It might be relevant if you're interested in a correlation of Vitamin D deficiency and anything having to do with the sun -- but that's not the correlation they're making here.
Is where the water comes from relevant to the correlation between where the water comes from and cholera? Tautologically yes.
Is where the water comes from relevant to the correlation between hydration and cholera? No.
The reason you need to be more specific is that these are completely different experiments:
If you want to see whether where the water comes from is correlated with cholera, you might observe everyone who goes to the well and everyone who collects rainwater for 30 days, and see whether more from the well group or more from the rainwater group develop cholera. Note: you aren't collecting any data in this experiment about how hydrated these people are.
If you want to see whether hydration is correlated with cholera, you might randomly select 30 people and measure how much water they collect every day for 30 days, and plot that against whether they develop cholera. Note: you aren't collecting any data in this experiment about where these people get their water.
"Is skin color correlated to vitamin D absorption from the sun?" and "Are vitamin D levels correlated with Covid19 infection and severity?" are two completely different questions and two completely different experiments. And to be clear, they aren't even connected by a common cofactor: "vitamin D absorption from the sun" and "vitamin D levels" are two very different things, which may not even be correlated in certain situations (that is to say, I may absorb very little vitamin D from the sun, yet have very high vitamin D levels--this could happen if I supplement or consume a lot of vitamin D in my diet). You cannot draw any conclusions about vitamin D absorption from the sun or skin color from this study, because it does not collect any data about those vitamin D absorption from the sun, and it does not collect any data about skin color.
> If you want to see whether hydration is correlated with cholera, you might randomly select 30 people and measure how much water they collect every day for 30 days, and plot that against whether they develop cholera. Note: you aren't collecting any data in this experiment about where these people get their water.
Your experiment is fatally flawed, due to the error described in your note. You would find cholera to be strongly associated with baseline hydration status--but that relationship is not causal, so your results would be misleading. An experiment not subject to this flaw would collect data about where people got their water, and control for that statistically.
Identifying confounding factors is a fundamental part of experimental design, because association alone is poor evidence of causation.
That's not an error, and the experiment isn't flawed. An experiment is only flawed if it fails to answer the question it's asking. You see this as a flaw because you're asking the question, "What causes cholera?" You're correct that the experiment doesn't answer that question, but that's not the question the experiment is asking. The experiment is asking, "Is where the water comes from relevant to the correlation between hydration and cholera?"
It's easy to say, "Well, that's the wrong question", with hindsight. You know that where the water comes from is a confounding factor because it's 2020 and we've known what the cause of cholera is for most of a century. But when you're designing experiments, you can only guess at what the right questions are, and you don't always know what the confounding factors are. Asking the wrong questions doesn't mean your results are wrong, it just means that you can't draw very broad conclusions from those results.
Broad conclusions are drawn from a large number of experiments. The broader the conclusion, the more experiments are necessary to prove it. Even your experiment which correlates hydration and cholera controlling for water source doesn't tell us what causes cholera. In 2020 we know that cholera isn't caused by well water--I have a Nalgene full of well water I've been drinking from since this morning and I don't think I'm in any danger. Is your experiment flawed? No! It just isn't asking the right question, yet! But it's a step toward the right question.
In practice, the way this would work is:
1. You do the hydration experiment and discover a correlation between cholera and hydration. Great! Maybe people are getting cholera because they drink too much water!
2. You design a new experiment where you try reducing water intake as an intervention. You randomly select 30 people to be in the experimental group and 30 people to be in the control. You tell the 30 people in the experimental group to drink 8 cups of water per day. Suddenly, your the benefits of drinking less water disappear. So now you know that there's a confounding factor which was present in the first experiment, but not the second.
3. You go back to inspect the first experiment and try to see what other variables might have been correlated with water consumption, and try to design an experiment that asks a different question. You might have to do this a bunch of times before you finally notice the well/rainwater difference, and decide to design an experiment asking about that.
That doesn't mean all your experiments were flawed! On the contrary, each of those experiments gave you a tiny bit of information which allowed you to eliminate a possibility or otherwise refine your questions until you found the exact right experiment to ask the right question.
> Identifying confounding factors is a fundamental part of experimental design, because association alone is poor evidence of causation.
Sure, but the only way you identify confounding factors is by doing experiments.
All of this is one huge unrelated tangent, because skin color isn't a confounding factor when the question is, "Is vitamin D level correlated to Covid19 infection and severity?"
In the specific example provided. Yes, of course it is.
Rainwater is close to pure water, drinking distilled water distroys cells in gut and causes diarrhoea, same as cholera does. Therefore this is a basic source of error in the experiment that should be controlled.
> Rainwater is close to pure water, drinking distilled water distroys cells in gut and causes diarrhoea, same as cholera does. Therefore this is a basic source of error in the experiment that should be controlled.
1. Where did you get this bizarre claim? They sell distilled water in stores for drinking. I've drunk it. It didn't cause me diarrhea, certainly not "same as cholera does". Cholera causes such severe diarrhea that people die of dehydration from it. There's some evidence that it slightly leaches minerals from your body, but that's a far cry from it causing such severe diarrhea that it might literally kill you.
2. Even if what you are saying were true (which is isn't), that would make it a confounding factor or regressor, not a source of error.
3. And even if it were a regressor, that doesn't invalidate the experiment. You can still draw a correlation between hydration and cholera, you just can't conclude that hydration causes cholera. Which is fine, because almost any experiment you do is going to require further experimentation to be able to establish causality.
Put another way, this experiment asks the wrong question, but asking a bunch of wrong questions is part of the process of science.
The cause of cholera has been known for about a century, so it's easy in hindsight to design an experiment that proves cholera is caused by microbes in fecally-contaminated water. But that's not how science works. In science you're asking questions where you don't know the answer, so you start by asking very narrow questions, like "Is cholera correlated with hydration?" and slowly expanding to questions like "Is cholera still correlated with hydration if you restrict hydration as an intervention?" or "Is cholera still correlated with hydration if you control for water source?" All of these are valid questions around which you can design valid experiments. You just have to understand that when you ask a narrow question you get a narrow answer. None of these experiments conclusively identify the cause of cholera--even the experiment you proposed.