Wow, perhaps this is my first real blog post since rebooting
And really one of my first blog posts ever.

I'm back from a really invigorating trip to CERN to give some statistics lectures to the CERN summer school. I played around with d3.js, a cool javascript library for doing data visualization in the browser. I also wanted to learn flask, a python based web micro-framework. I got my first non-trivial flask+d3 example going and put it up here on GitHub. Just to make my self super-trendy, hipster, data scientist guy I also ditched emacs for Sublime Text 2. This is actually a huge change given that I'm a power-user in emacs and can do some pretty cool keyboard macros.

While I was there I met with Salvatore Mele and Sünje Dallmeier-Tiessen from INSPIRE and CERN open science initiatives. Some exciting things on the front of presenting LHC results, perhaps publishing 2-d profile likelihoods (slices of Higgs likelihoods in thousands of dimensions). It's amazing how the model independent (cross-section)_i x (branching ratio)_j contours (for i={VBF+WH, ggF+ttH} and j={WW, ZZ, γγ})

are sufficient statistics for for a number of benchmark-model coupling measurements (eg. kV-kF, kg-kγ, λ_WZ, etc.)

In addition, I just finished  the process of uploading data from a phenomenology paper I worked on a few years ago where we did a big MCMC (Markov Chain Monte Carlo) scan of the CMSSM (Constrained Minimal Supersymmetric Standard Model). So far I've uploaded it to both Figshare and to the Harvard Dataverse Network. I think I'm going to go with Dataverse though Figshare has been very nice. I'm choosing Dataverse mainly b/c of attention to meta-data and ability to launch a branded instances of the site. Though it did take me a hours to get the upload to work and it only took about 10 seconds to upload it on Figshare. Also, I want a nice API to tabular data so that I can link it with d3.js... neither have it, but FigShare's seems further behind (surprisingly b/c they are much more "2.0"-ish).

So I've released my first dataset today (July 12, 2013) and it's here:

Cranmer, Kyle; Allanach, Ben; Lester, Christopher; Weber, Arne, "Replication data for: "Natural Priors, CMSSM Fits and LHC Weather Forecasts"", hdl:1902.1/21804 UNF:5:bvw2MoOATltX0yfGjF9hJQ== V1 [Version]

If you go to it and click on one of the TabularData entries you can even do some minimal analysis from the web. Like this


I wish they would give me a little gist-style embed like this one:

Anyways, that's a lot of playing around with new things of late. I hope to see the DataVerse handle associated to the INSPIRE record soon.


comments powered by Disqus