The one problem here is, it would be really difficult - I would argue impossible - to demonstrate any effect in a convincing way. It's impossible to run a contr

The one problem here is, it would be really difficult - I would argue impossible - to demonstrate any effect in a convincing way. It's impossible to run a controlled experiment, and it's impossible to tease out and pin down all the confounding variables. It's what we referred to in one class I took as "fundamentally unanswerable questions." (Yes, the acronym was deliberate.) Worse, with the way standard statistical methods have been laid out, if you ignore the fuqedness of the question and go looking for correlations anyway, you are virtually guaranteed to find all sorts of them. About 1 in 20 things you think to look at will yield a statistically significant result. By definition. The classic XKCD on this effect is https://xkcd.com/882/ https://xkcd.com/882/