> Yes, so there will be a strong correlation between tight control over flouride in drinking water (e.g. clamping it to 1mg/L or so versus not deflouridating at

> Yes, so there will be a strong correlation between tight control over flouride in drinking water (e.g. clamping it to 1mg/L or so versus not deflouridating at 10mg/L levels) and a wealthier environment with good infrastructure. I think it's just so hard to "weight" for the differences here. Except that they used models in many studies, likely with factors for wealth, area, etc. This attempts to control for those factors, meaning they no longer have their intuitive effect on outcomes. Now, I say "attempts", because it corrects for some measurement of wealth and area, not the full underlying structure. You're basically left with an "empty" correlation between IQ and fluoride in the data, but useless for causation conclusions. By "empty" I mean that you have attempted to correct for all known factors like some measurement of wealth, but there are always still many unknown underlying factors that might cause issues. That's what randomization solves. Also note that many of the source studies are from China. This might sound odd to people outside academics, but due to many factors you should generally judge studies from most Chinese universities as less trustworthy than from other countries.