The data comes from Quest labs that allowed the researchers to use the unique patient ID to associate SARS-CoV-2 PCR lab results with 25(OH)D blood test results. The dataset includes 190K patients.
This is pure correlation, as Deva Boone’s excellent blog clearly explains.
However, a few days ago there was on the HN front page a link to what appears to be causal effect - Covid patients who were given vitamin D On admission did better than those who didn’t. That study had some limit and they had a couple of uneliminated confounders - but the body of evidence does hint that proper vitamin D supplementation is likely beneficial in fighting Covid.
Massive effect size which somewhat compensates for the small N in this seemingly well-designed study.
I suppose inferring causality still requires a leap of faith...
If someone said they fed a 3 month old mouse a 0.1 mg dose of Chemical X and it died immediately. Hard to argue that sample size of 1 means the result is worthless anecdote. The observability is pretty good.