Re:Vitamin D uncorrelated with Covid severity in those of Eu...
by legerdemain on Wednesday January 20, 2021 @03:48PM
(#25851781)
Sure, and another way to get a zero correlation is if you arrange all the data points evenly around a circle.
Are you seriously suggesting that authors who took the energy to get a manuscript reviewed and published in a BMJ publication wouldn't examine a scatter plot or check the estimate of a quadratic trend that their analysis software almost certainly outputs?
Re:Vitamin D uncorrelated with Covid severity in those of Eu...
by curiousgal on Wednesday January 20, 2021 @03:56PM
(#25851873)
My point pertains to the conclusion the public might draw from the study. Just because Vitamin D is uncorrelated with Covid severity does not necessarily mean that they are not related. Same way Y depends on X but is uncorrelated with it.
Re:Vitamin D uncorrelated with Covid severity in those of Eu...
by legerdemain on Wednesday January 20, 2021 @04:04PM
(#25851963)
Yes, and my point is that for your point to be relevant, the authors would have to exhibit a negligence that borders on the fantastical.
No one would look at a significant quadratic trend in the data and say, "Nope, not gonna publish that juicy nugget!"
Re:Vitamin D uncorrelated with Covid severity in those of Eu...
by whoisburbansky on Wednesday January 20, 2021 @03:25PM
(#25851502)
The correlation here averages out to zero because X can be negative but Y can't, right, what are you trying to show here?
Re:Vitamin D uncorrelated with Covid severity in those of Eu...
by klmadfejno on Wednesday January 20, 2021 @03:29PM
(#25851575)
Not only that, but while there's no LINEAR correlation, there's a perfect non linear function that any half-credible scientist would identify.
Re:Vitamin D uncorrelated with Covid severity in those of Eu...
by whoisburbansky on Wednesday January 20, 2021 @03:32PM
(#25851603)
Agreed, but I still don't see how that's relevant to the Vitamin D/Covid severity correlation question. Presumably you aren't modeling either of the associated parameters with random variables that can be negative, and so even if one were a non linear function of the other, the linear correlation wouldn't be zero?
Re:Vitamin D uncorrelated with Covid severity in those of Eu...
by klmadfejno on Wednesday January 20, 2021 @03:41PM
(#25851701)
shrug. You could imagine, for example, that too little Vitamin D is bad, and too much is bad, and arrive at a similarly (flipped) nonlinear fit with negligible linear correlation
Re:Vitamin D uncorrelated with Covid severity in those of Eu...
by curiousgal on Wednesday January 20, 2021 @03:29PM
(#25851566)
It's a standard example of why non-correlation doesn't mean independence.
Let Y be equal to X*X.
What would the correlation between X and Y be?
Edit: point being that non-correlation doesn't mean independence.