Comment by Invictus0

> I think was just an overfit model producing nonsensical results. That kind of statistically-significant-but-nonsensical parameter is exactly what Freedman's Paradox describes.

You just completely made this up. There isn't even evidence that a "model" exists or was fitted to.