[expression1sh]

The actual study in question:

https://www.jsad.com/doi/pdf/10.15288/jsad.23-00283

I'm not going to add any commentary on the paper, other than to say that the risk ratio for all-cause mortality for moderate drinkers in the six high-quality papers they identified is indistinguishable from zero.

In other words two to three drinks per week has not been identified net causing any harm, at least in epidemiological studies.

[pengaru]

"all-cause mortality" vs. "any harm" is quite the leap

[graemep]

People assume correlation implies causation when it strengthens their case.

They assume correlation does not imply causation when it weakens their case.

Not only true for alcohol consumption. It seems to be true for pretty much anything when people use graphs showing correlations about anything complex.

Its particularly bad with time series data where simple correlations are essentially meaningless (yes, there are more advanced methods for dealing with this).