Statisticians' Call To Arms: Reject Significance And Embrace Uncertainty!

[KHashmi317]

Statisticians' Call To Arms: Reject Significance And Embrace Uncertainty!

"A recent study that questioned the healthfulness of eggs raised a perpetual question: Why do studies, as has been the case with health research involving eggs, so often flip-flop from one answer to another?

...

An entire issue of the journal The American Statistician is devoted to this question, with 43 articles and a 17,500-word editorial ..."

Article and NPR podcast here:

https://www.npr.org/sections/health-shots/2019/03/20/705191851/statisticians-call-to-arms-reject-significance-and-embrace-uncertainty

[TomBAvoider]

Right. This doesn't mean that anyone (that I know of) rejects the reporting of p-values, just that there is no longer any hard line with a specific number that if you fall over in one direction it's called "significant" and if the other way "not significant". Which to be fair was always somewhat arbitrary - why 0.05, and what happens that's so dramatic at 0.049 or whatever. The specific number and label was developed for convenience. What happened is that then this number became something that was sometimes gamed in order to push into significance thus distorting the integrity of some studies - take away a specific number you have to hit, and you take away the motivation to game and can report fairly... is how the thinking goes. It became a negative when journals would reject or not publish papers that didn't hit that magic number, and that way potentially valuable studies were not accessible. 

In other words, this could be a positive development. It doesn't mean that the very concept of significance has been abolished, just that now the number is reported and we all can decide whether we think it's significant or not - sometimes something that approaches significance in the old sense of 0.05 though not formally reaching it can nonetheless be a valuable pointer as to a direction for further exploration, so there is value in numbers other than 0.05, and there isn't a need to obsess over an arbitrary one like 0.05. I think it's all good science-wise. Now, how that will play out in the *reporting* of science is a whole other ball of wax, and here I do worry slightly that marginal results will now be trumpeted from the rooftops... but then again, I remind myself that even under the supposed tyranny of 0.05 as cutoff for significance,  the pop scienc press would still trumpet marginal results that did not reach that magic number - so it's not like some kind of virginity has been lost - maybe the only danger is that now nobody can say "significant" with authority, and everyone must decide for themselves leaving more room for hype. I'm OK with all this, as long as the p-value does get reported and is not ignored altogether.

[Clinton]

I’m wondering how bad eggs ‘really’ are.  2 eggs have approx 373mg of cholesterol not to mention fairly significant amounts of choline.   The choline, minerals and amino acids profile make eggs a nice addition to a vegetarian diet ...

[mikeccolella]

Different people react differently to egg consumption wrt tmaos. It appears that the microbiome is the factor. 

[Todd Allen]

Eggs are my top source of protein but I produce my own.  I wouldn't touch most commercially produced eggs although if I lived next to this guy I'd buy eggs from him since he feeds and raises his chickens similarly to how I do it.

How Karl Hammer Feeds 600 Chickens (Without Grains)

[mccoy]

The nature article is interesting

By the way, as most of us know another aspect which should be evaluated is the absolute risk or attributable risk of a particular regimen together with the relative risk which is usually depicted in metanalyses