For discussion of the primary matter of this thread, but the stubborn dissagreements can remain there.
Follow along with the video below to see how to install our site as a web app on your home screen.
Note: This feature may not be available in some browsers.
Cool digitization tool. Yes, I believe your r^2 =0.0884 a lot more than their r^2 = 0.98. And yet this is a peer-reviewed article in a prestigous journal (PNAS - an initialism, not an acronym). I can't help but wondering if we're missing something somehow. Too bad we can't ask the authors to explain how the heck they came up with their number.IRStuff said:I suspect they were clutching at straws. Linear regression is sensitive to extrema, so BMI*yr potentially moves some datapoints around enough to get the 0.98 regression coefficient. Nevertheless, the fit is pretty awful, and correlation does not equal causation.
Even worse, I think they fudged the data. I digitized the superspreader datapoints and got a similar regression line, but R^2 = 0.0884, which is essentially no correlation. Image I used and spreadsheet attached. I used to digitze