Paywalled, can't see. But if it's empirical, a positive correlation on top of a negative one, would be undistinguishable from no correlation. This in fact may b

Paywalled, can't see. But if it's empirical, a positive correlation on top of a negative one, would be undistinguishable from no correlation. This in fact may be very common, but the bottom line is what the sum of the effects is rather than the individual forces. Suppose the same can be said for alcohol and relaxation, if we see evidence of more heart attacks, it means the negative effect is bigger than the (hypothetical) positive.