The key trick here is that they do a statistical test for interaction [0] which basically asks "does one variable influence the effect a second variable has on

The key trick here is that they do a statistical test for interaction [0] which basically asks "does one variable influence the effect a second variable has on the outcome?". This doesn't immediately trip my BS sensor, but I haven't read past the abstract. This could easily be tripped up by poor data or bad modelling. 0. https://en.wikipedia.org/wiki/Interaction_(statistics) https://en.wikipedia.org/wiki/Interaction_(statistics)