The first comments follow the discussion at this blog link:
A few things are wrong here with some remarks (and misunderstanding) I point out one misunderstanding: the bump is a tail moving slowly as per Monte Carlo.
“The bump is clearly a Gaussian so we can’t say it’s a tail that’s moving slowly because of simulation”
1. A Gaussian is never a tail and a tail is never a Gaussian. Gaussian can happen on top of tails, they just like to sit anywhere they are created. If an Australian baby is borne in woods it may just sit on the top of a Kangaroo’s tail.
2. To say the bump is not predicted by standard model is to say this bump is not included in Monte Carlo. SO why would MC have a say here. It’s all experimental data.
3. A 3-sigma effect is not something we wish “go away” (as pointed out) It could be real, we need more data and more detailed study. Period. It’s not just that a 3-sigma effect would grow into a 5-sigma effect; it’s also that it’s never going to go below. SO why would anyone say 3-sigma is not important? It’s just that most people would put it under carpet and never bother again.
4. This bump could be an enigma, playing hide and seek with “slice and dice” not necessarily getting cut off.
Now I would like to point out that the bump is actually a double Gaussian not a single. See, one data point is off from the Gaussian fit at the same level of the peak of the Gaussian. SO it’s better fitted with two Gaussians, which would just mean that the bump has two resolutions, a usual detector effect. But there is also one more thing to note here the peak of the green; it’s slightly to the left of the mean of the red peak.
And about discovering a new fundamental force of nature, are we prepared theoretically or experimentally for such a scenario? I had pointed out we may have more 4 fundamental forces of nature, in an article of mine 3 years ago. That’s the real power of LHC.
These comments below follow the discussion on facebook, see the plots posted on top.
is not really super-convinced by this CDF excess on the side of a falling distribution, given that the data and background model don’t fit at the top of the distribution. If I shift everything over one bin before background subtraction, is there any excess?
Herbi Dreiner
Plus if you look on the web page you’ll see that the electron and muon data are quite different, which shouldn’t be for a W+ jets.
Naveen Gaur
I suppose this (difference of muon and electron) was already there earlier too (may be due to detection efficiencies) !! There are two papers today on arxiv attempting solution of the CDF. They attempted resonant production, I am puzzled as CDF has clearly stated that this possibility is not there !!
Physicist Matthew Strassler
muon/electron differences are common (you will see this all over the LHC results if you look) — the issue is whether the difference in detection efficiencies are understood by the experiments or not.
Physicist Matthew Strassler
Hey, Joanne, could you get Jared to repost Figures 1b and 1d from the paper with the resonant contribution subtracted? It should show that not only has CDF discovered a particle, it has also discovered a ghost.
Herbi Dreiner
Sorry, I disagree. After correcting for efficiencies the muon and electron Xsections should be the same. It’s supposed to be a W! But they are just barely consistent at 1 sigma. I find that dubious.
JoAnne Hewett
Jared?? We’d all love to see the plots!
JoAnne Hewett
Ah – I see Yuri posted it!
Naveen Gaur
But I think CDF paper says that resonant contribution (M_{l j j MET}) is consistent with SM shape. So probably there is no resonent contribution.
Manmohan Dash
The fit is a little whacky, but shifting by 1 bin you lose the 3-sigma effect, you gotta be kidding. There are 8 bins of effect here. Since the fit is a little off from data points so also is the bump. The bump should be fiitted to an additional gaussian…http://blogs.discovermagazine.com/cosmicvariance/2011/04/06/anomalies-at-fermilab/comment-page-1/#comment-158561
Physicist Matthew Strassler
No, Manmohan, this is not true. The point has been nicely illustrated in http://cmsdoc.cern.ch/~ttf/CDFDiJetScale/AnimatedDijet.gif [who made this plot?!] Notice a 2 percent shift in energy scale makes the new “bump” look a lot less impressive — and the fit is still good. And by 3 or 4 percent it is almost gone — and the only problem with the fit is at very low energy, where there can be subtleties with jets.
Herbi Dreiner
As far as I know the plot was made by Adam Falkowski.
Physicist Matthew Strassler
I am not convinced of that.
Manmohan Dash
1. The peak is sitting on a flat surface where as the bump is sitting on a highly curved surface. 2. The top left plot is showing clearly the bump is not a wish-bean. Rather I am more uncomfortable with the fit. The fit isnt sitting where …it must plus as I noted the bump may need an extra Gaussian. 3. On the animation which is quite fast and quite hard on the eye, its of not much use except the numbers you quoted. Let’s say 2% shift what does it do to the peak? (rather than the bump) Lets say 4% where you say the bump is washed away, whats the effect on the peak? When there is a shift in energy is every parameter floated or some are fixed. Many questions, but at the end the bump is not small (3.2 sigma it is, it lives only 0.1% off, thats the chance we will have to disregard this bump or we have something growing here). The real way ahead would be to double or triple the data without changing a damn cut or optimization, floating or fitting.
Manmohan Dash
On the animation [right plot] the top one is meaning less. One way to see that would be to see when there is an energy shift on the top plot the peak is growing when the bump is reduced and on the bottom plot the peak is reduced with the bu…mp reduction. That’s because the bottom has background bins where the bump events are accomodated when the energy is shifted. (When the energy shift is “actuated” where does all the events of the bump go?) On the bottom plot the peak is decreasing so they are going to the background bins. WHY are WE shifting the energy ?? WE should just divide the sample into 10 samples and fit each ??
"Various Musings" Journal of IPH, ajwad of MDashF 