4 ms·
I haven't dug into this, and, for the record, I do think vitamin D is likely an important factor here. But I find the two charts to be very worrisome. Their cl
by klmadfejno 6y ago
I haven't dug into this, and, for the record, I do think vitamin D is likely an important factor here. But I find the two charts to be very worrisome.
Their claim of a very high p value and significant effect is VERY strongly influenced by a single outlier. It's not even close. On this unlabeled graph, I think x axis is vitamin D levels and y axis is deaths per million. They've got ~16 data points in the range of 0-10 deaths, one point at ~20, and one at 65. If you removed the 65 point the line would get a hell of a lot less convincing.
- zebrafish 6y agoThis study is not exactly rigorous. There doesn't appear to be any controls for age or health risks or anything else. Their methods are literally to take worldometers data and vitamin D concentration data per country from the European Calcified Tissue Society and plug those values into a t-test calculator. When HN was in a tizzy over p-values a year or so ago, this is what people were concerned about. It's below .05, so what? Not saying there is or isn't an affect on covid by vit D, but this isn't the study to prove it.
- namdnay 6y agoWait a sec... they're taking their data at country level? How on earth is that supposed to mean anything for anywhere apart from Belgium and the like?
- marcosdumay 6y ago> Their methods are literally to take worldometers data ... and plug those values into a t-test calculator. I have seen more papers like this recently. It seems to have become a popular format in hypothesis discovery, usually followed up by others with better methodologies. They become a problem if the popular press takes the results to publish out of context, or if they enter meta studies.