Re:Japanese scientists develop vaccine to eliminate cells be...
by derefr on Friday December 17, 2021 @01:31PM
(#29596221)
> It basically has infinite confounding variables.
Yes, but: Law of Large Numbers. The great thing about retrospective data-collection studies is that you can potentially get hundreds of thousands of histories as datapoints. With those numbers, basically every confounder washes away, other than the selection effect of "what kind of person frequents this hospital rather than some other hospital." And you can get rid of that confounder, too, if you meta-analyze studies done in hospitals in very different areas.
Yes, but: Law of Large Numbers. The great thing about retrospective data-collection studies is that you can potentially get hundreds of thousands of histories as datapoints. With those numbers, basically every confounder washes away, other than the selection effect of "what kind of person frequents this hospital rather than some other hospital." And you can get rid of that confounder, too, if you meta-analyze studies done in hospitals in very different areas.