Inspired by the Higgs, a step forward in open access

[originally posted on Quantum Diaries]

The discovery of the Higgs boson is a major step forward in our understanding of nature at the most fundamental levels. In addition to being the last piece of the standard model, it is also at the core of the fine tuning problem -- one of the deepest mysteries in particle physics. So it is only natural that our scientific methodology rise to the occasion to provide the most powerful and complete analysis of this breakthrough discovery.

This week the ATLAS collaboration has taken an important step forward by making the likelihood function for three key measurements about the Higgs available to the world digitally. Furthermore, this data is being shared in a way that represents a template for how particle physics operates in the fast-evolving world of open access to data. These steps are a culmination of decades of work, so allow me to elaborate.

 

First of all, what are the three key measurements, and why are they important? The three results were presented by ATLAS in this recent paper. Essentially, they are measurements for how often the Higgs is produced at the LHC through different types of interactions (shown above) and how often it decays into three different force carrying particles (photons, W, and Z bosons). In this plot, the black + sign at (1,1) represents the standard model prediction and the three sets of contours represent the measurements performed by ATLAS. These measurements are fundamental tests of the standard model and any deviation could be a sign of new physics like supersymmetry!

Ok, so what is the likelihood function, and why is it useful? Here maybe it is best to give a little bit of history. In 2000, the first in a series of workshops was held at CERN where physicists gathered to discuss the details of our statistical procedures that lead to the final results of our experiments. Perhaps surprisingly, there is no unique statistical procedure, and there is a lot of debate about the merits of different approaches. After a long discussion panel, Massimo Corradi cut to the point

It seems to me that there is a general consensus that what is really meaningful for an experiment is likelihood, and almost everybody would agree on the prescription that experiments should give their likelihood function for these kinds of results. Does everybody agree on this statement, to publish likelihoods?

And as Louis Lyons charred the session...

Any disagreement? Carried unanimously. That's actually quite an achievement for this workshop.

So there you have it, the likelihood function is the essential piece of information needed for communicating scientific results.

So what happened next? Well... for years, despite unanimous support, experiments still do not publish their likelihood functions. Part of the reason is that we lacked the underlying technology to communicate these likelihood functions efficiently. In the run up to the LHC we developed some technology (associated to RooFit and RooStats) for being able to share very complicated likelihood functions internally. This would be the ideal way to share our likelihood functions, but we aren't quite there yet. In January 2013, we had a conference devoted to the topic of publishing likleihood functions, which culminated in a paper "On the presentation of LHC Higgs results". This paper, written by theorist and experimentalists, singled out the likelihood associated to the plot above as the most useful way of communicating information about the Higgs properties.

The reason that these specific Higgs plots are so useful is that more specific tests of the standard model can be derived from them. For instance, one might want to consider beyond the standard model theories where the Higgs interacts with all the matter particles (fermions) or all the force carrying particles (vector bosons) differently than in the standard model. To do that, it is useful to group together all of the information in a particular way and take a special 2-d slice through the 6-d parameter space described by the three 2-d plots above. To the left is the result of this test (where the axes are called κ_F and κ_V for the vector bosons and fermions, respectively). What is special about this plot is that there is an overlay of the original ATLAS result (filled contours) and those reproduced from the official ATLAS likelihood functions. While my student Sven Kreiss made the comparison as part of a test, anyone can now reproduce this plot from the official ATLAS likelihood functions. More importantly, the same procedure that was used to make this plot can be used to test other specific theories -- and there are a lot of alternative ways to reinterpret these Higgs results.

Great! So where can you find these likelihood functions and what does this have to do with open access? I think this part is very exciting. CERN is now famous for being the birthplace for the world wide web and having a forward-looking vision for open access to our published papers. The sheer volume and complexity of the LHC data makes the conversation about open access to the raw data quite complicated. However, having access to our published results is much less controversial. While it is not done consistently, there are several examples of experiments putting information that goes into tables and figures on HepData (a repository for particle physics measurements). Recently, our literature system INSPIRE started to integrate with HepData so that the data are directly associated to the original publication (here is an example). What is important is that this data is discoverable and citable. If someone uses this data, we want to know exactly what is being used and the collaborations that produced the data deserve some credit. INSPIRE is now issuing a Digital Object Identifier (DOI) to this data, which is a persistent and trackable link to the data.

So now for the fun part, you can go over to the INSPIRE record for the recent Higgs paper (http://inspirehep.net/record/1241574) and you will see this:

If you click on HepData tab at the top it will take you to a list of data associated to this paper. Each of the three entries has a DOI associated to it (and lists all the ATLAS authors). For example, the H→γγ result's DOI is 10.7484/INSPIREHEP.DATA.A78C.HK44, and this is what should be cited for any result that uses this likelihood. (Note, to get to the actual data, you click on the Files tab.) INSPIRE is now working so that your author profile will not only include all of your papers, but also the data sets that you are associated with (here's anexample in it's early form based on my ORCID ID).

Now it's time for me and my co-authors to update our paper "On the presentation of LHC Higgs results" to cite this data. And next week, Salvatore Mele, head of Open Access at CERN, will give a keynote presentation to the DataCite conference entitled "A short history of the Higgs Boson. From a tenth of a billionth of a second after the Big Bang, through the discovery at CERN, to a DataCite DOI".

I truly hope that this becomes standard practice for the LHC. It is a real milestone for the information architecture associated to the field of high energy physics and a step forward in the global analysis of the Higgs boson discovered at the LHC!