Fermi Lab May Have Discovered New Particle or Force

schleprock63 writes "Physicists at Fermi Lab have found a 'suspicious bump' in their data that could indicate they've found a new elementary particle or even a new force of nature. The discovery could 'be the most significant discovery in physics in half a century.' Physicists have ruled out that the particle could be the standard model Higgs boson, but theorize that it could be some new and unexpected version of the Higgs. This discovery comes as the Tevatron is slated to go offline sometime in September."

Comments

Do they account for hypothesis-mining?

[Rich0]

When I read things like "In about 250 times more cases than expected, the total energy of the jets clustered around a value of about 144 billion electron volts" I get nervous.

This is like saying that in a series of 1M coin tosses the sequence HTTHHTTTTHHTTHHH came up 100x more often than would be expected by chance. Does that mean that any particular sequence of 8 tosses should come up 1/65536th of the time, and this one came up 1/655th of the time, or does it mean that some random sequence of results should come up 100% of the time in a random series of 16 coin tosses, and we happened to pick the random series that came up the most often in that particular set of data?

If I mine a big set of data against 100 random hypothesis I'll be able to find about 5 that I can show to be true with 95% confidence, despite the fact that there is nothing really going on.

The real test is to come up with the hypothesis first, then collect the data.

Now, these guys are probably smart, and hopefully control for this. If you want to test for 100 hypotheses and REALLY have 95% confidence, then you need to target a confidence of 1-0.05^100 for each test - at least that is how I see it (being a complete novice at statistics).

Re:Do they account for hypothesis-mining?

[radtea]

The real test is to come up with the hypothesis first, then collect the data.

That's not the way the vast majority of science is done. Popper was a philosopher speaking in ignorance (but I repeat myself).

The challenge for these guys is not in the hypothesis testing, but in the cuts. You have to come up with some set of criteria for selecting "good" events in complex detectors of this kind. There is always a degree of arbitrariness in how you do that, and there have been cases in the past (the so-called 'GSI particle') where people tweaked and tuned multi-dimensional cuts to maximize peaks in the data.

In the present case it is clear their cuts are physics-based--they are described in the paper--and that the peak structure is consistent with the resolution one would expect (the GSI particle required some very weird physics to make the narrow peak widths plausible.)

However, the peak is also precisely in the region where their background spectra are varying most rapidly, and this is a huge red flag. It makes them sensitive to any number of minor mis-calibrations. It does NOT mean the phenomenon is not real, but if I had to make a bet on it being physics beyond the standard model or an instrumental artefact, my money would not be on new physics.

A useful link

[jd]

...to the paper, as opposed to the commentary by PopSci on the article written by NYT by someone who really didn't know what the hell they were talking about.

Re:A useful link

[Anonymous Coward]

And a link to the lecture set to go live in an hour on it:
http://vms-db-srv.fnal.gov/fmi/xsl/VMS_Site_2/000Return/video/r_livelogicindex.xsl?&-recid=573&-find=
(posted AC, you dirty karma whore)

Re:A useful link

[jd]

Nobody views under a 2 any more, so if you want a link to be seen you have to post non-AC. No choice. Too b. noisy with trash talk otherwise. I'd not have seen your link at all if I didn't have a habit of expanding hidden replies on the offchance they're important.

(And because very few people mod ACs - why bother, it won't alter their karma - important AC posts often vanish into the ether.)

Re:Sad to lose the Tevatron

[The_Wilschon]

FWIW, earlier drafts of the paper were much more sensationalistic than the final draft that the collaboration approved. A large contingent of the collaboration, myself included, would have removed our names from the paper if it had done something as insane as claim discovery of a new particle. So, we specifically pushed to make the paper more scientifically honest and less effective as a "ploy to keep the funds flowing." That said, the NYT article and all the other mainstream news reports on the issue are far, far more sensationalistic than anything the analyzers ever even considered producing...

Some interesting things to note:

• This search was done with a very tight event selection designed to get a relatively pure diboson sample. Loosening up the selection increases the number of data events involved in the analysis by (IIRC) about a factor of 8, and in this looser sample, the significance of the bump decreases to about 1 sigma, which is wholly uninteresting.
• The feeling among the members of my particular group (one of the member institutions of the CDF collaboration) is that this is a very interesting result, but that it should be interpreted more as exposing the difficulties of / our inability to model the very large W+jets background accurately; the Monte Carlo generators are simply insufficient or are slightly incorrectly tuned. We do not really feel that this is likely to be an indication of new physics at all.

So, long story short, there is certainly something here to be interested in. Both the theorists who write the Monte Carlo generators and the experimentalists analyzing data from the LHC experiments are paying close attention to this result, as it affects their work. We will know more after further study and work, both to improve the Monte Carlos and to look for similar effects in the ATLAS and CMS data.