Hacker News Reader

Comment by ced | Hacker News Reader

3.4 sigmas is the within-model uncertainty - that is, given the assumptions about their instruments and methodologies, the result stands at 3.4 sigma, or about 1/1000.

I think that the uncertainty about the model and assumptions is already larger than 1/1000, especially given the recent failed "faster-than-light neutrinos" LHC experiment. So while 5 sigmas would be nice, I would assume that the physicists over there are busy double-checking everything, and trying to reproduce the result, rather than improve their "sigma-score".

Can any physicist confirm? I'm not all that familiar with how these things work. Furthermore, if they perform once more the experiment and get the same 3-sigma result, wouldn't that compound into a >3-sigma total?

View on HN

The faster-than-light neutrinos probably got explained by a loose cable, sorry. :) http://arstechnica.com/science/2012/02/faster-than-light-neu...

As to the double-checking, the analysis behind this result used the whole data set collected by the BaBar experiment at the Stanford Linear Accelerator (SLAC). So that's years of data, and the experiment hasn't been running since 2008. The question of course is, can some other experiment produce the same results. As this is B-physics I would guess LHCb, one of the 4 main experiments at CERN's LHC, might be able to do this, but I really don't know for sure.

If you can repeat the same experiment and get another 3 sigma deviation from the Standard Model predictions then yes, you could combine the results to get a >3 sigma total (though they don't simply just add up to 6).

The faster-than-light neutrinos probably got explained by a loose cable, sorry. :)

Right:) That's exactly what I was referring to: the uncertainty about the model/experimental setup is larger than the N-sigmas uncertainty reported by the model and instruments.

Right, sorry, I misunderstood. You are right and this was explained well by yk a couple of posts below.

Except the scientists behind the loose cable up front said "we don't understand these these results, we think they are false, please help us figure out why". At no point were they published for legitimate consideration.

There are different sources of errors, usually categorized as statistical and systematic. The statistical error will decrease with higher statistics but the systematic error will not. An example for a systematic error would be the error of the energy calibration. In the paper they claim for one of the measured ratios R(D)=0.440+-0.058 (statistical) +- 0.042 (systematic) and the result would at best get about sqrt(2) as good if you would get rid of the statistical error. (Assuming the errors are uncorrelated they combine as the root of the squares).

I wouldn't put much stock in the faster-than-light neutrino result. That was never peer-reviewed, contradicts the results of other experiments on the same path, and led to two of the leaders of the OPERA project being ousted:

http://latimesblogs.latimes.com/world_now/2012/04/faster-tha...

It was a lot of media hype over a preprint of preliminary results that have never been confirmed.

It was a cable. Just that. They published the data knowing it was wrong - they asked fellow scientists to explain why.

Reply on news.ycombinator.com