A Grauniad post from Jon Butterworth – I’m becoming a fan, one of the more down to earth physicists I’ve met.
But, it’s another one of those nagging doubts I have. I get the 5-sigma stats, the limits to the assumptions of normal distributions, and even the subconscious Bayesian correction, but I can’t help feeling the focus on error relative to your thesis is merely reinforcing – reifying – the objectification of your own thesis. It’s about error “relative to” your set of assumptions. The smaller this error the greater the significance of some bigger error in your model / hypothesis / assumptions built into the measurements you’re seeking as well as the results you’re finding. Your ability to achieve “perfect” results is greater the more your boundary conditions constrain what you can look for.
I suspect this is tied up in the “look elsewhere” idea he mentions – which I don’t really get, yet. More reading. Sigh!
Interestingly even in the title / intro / abstract – Lyons agenda seems to be aligned with some of my issues. The rationalisation and (resource) justification aspects of large physics projects might create some self-fulfilling relationship between theory and results. Intriguing.
[Post Note : and the following day a post taking supersymmetry possibilities seriously at 2.6-sigma. Previous supersym ref here. Also ephemeral points on twitter – that big physics is using 5-sigma just because it can, given how many events it can poll. Reinforces my nagging fear that the stats are a red-herring, missing (obscuring) something of more explanatory value, more real than the standard model.]
[Post Note : Opportunity for a (real, attributable) Einstein quote
“Since the mathematicians have invaded the theory of relativity,
I do not understand it myself anymore.”
(Albert Einstein : Philosopher Scientist (1949) edited by Paul A. Schilpp)]