[DISCLAIMER: NONE OF THESE RESULTS SHOULD BE TAKEN AS BLESSED OR ACCEPTED IN ANY WAY. EVERYTHING PRESENTED HERE IS MERELY A WORK IN PROGRESS AND BASED ON MONTE CARLO SIMULATION, NOT DATA. THIS SITE EXISTS FOR MY PERSONAL USE ONLY AND NOT REPRESENTATIVE OF THE CURRENT STANDING OF PARTICLE PHYSICS RESEARCH.]Contact Information
Email: jnett@wisc.edu
Recent Work I've joined the H->WW group to conduct a Trilepton analysis for my thesis.
Here are the discriminating variable candidates for the WH sample, meaning the Z-Peak is removed. That and the other analysis cuts are described on the introductory slide. It looks to me like the best candidates are MET, deltaR and deltaPhi, and to a lesser extent transverse mass.
For ZH, it looks like mass will be the key discriminating variable here. Particularly,
1. Inv. mass of 3rd lepton + met + jets
2. H_T: leptons + met + jets
3. m_T: 3rd lepton + met
4. m_T: opposite-signed close (by deltaR) leptons + met + jets
5. NJet
WH120 Discriminating Variables ZH120 Discriminating Variables WH160 Discriminating Variables ZH160 Discriminating Variables WH190 Discriminating Variables ZH190 Discriminating Variables
Also, here is a comparison of just the signal curves for 110 - 200 GeV:Signal Comparison
Compare Drell-Yan and Z\gamma .
(15 October 2009) Here are my first neural net results. In the "No Z-Peak" region, it appears that the Z\gamma background is discriminated against very well. However, in the "In Z-Peak" region, I'm a little surprised the WZ background is not discriminated against better than it is given the distributions of many of the variables I chose. I'm still playing with them.
(21 October 2009) Compare ttbar with and without matched letpons
(21 October 2009) Here are some notes I have compiled concerning how to setup the NeuroBayes software. I hope to later add a section with tips on how to train the neural network successfully--once I figure that out more for myself.
NeuroBayes Notes
Older Work
Stntuple Guide: I spent a frustrating couple of months figuring out how to build the basic framework for a module that inherits from TStnModule, compiles, runs, and then expand this into a collection of modules constituting an analysis that fills personal ntuples and analyzes them for some purpose (a background analysis in this case). More generally, the experience focused my awareness of the necessity for streamlined communication channels in an organization that covers many disparate projects that are united by common goals and shared tools. As physicists and physics students, I believe we should distinguish between our objectives (physics research) and the tools we utilize for the purpose of that research (programming frameworks, for example). I wrote this Stntuple guide as a result of my insistence that it should already have existed. What took me a couple months to figure out (and write up here) should take the new graduate with only basic understanding of Linux and C++ days to complete, and the more seasoned research just hours, with this in hand. Some may argue there exists benefit from the experience of 'figuring it out for yourself' and this is certainly true when speaking of math and physics problems, but is a sentiment applied quite out of context here. Figuring out the basic analysis structure for Stntuple is far less a matter of logical deduction and inspection and for more a matter of arbitrary minutia that one must simply be told. For this reason, detailed step-by-step instructions to get started are a necessity.
It would also be a mistake to think the value of such documentation is simply a matter of preventing graduate student headaches. Far more importantly, this is a matter of efficient use of research resources. If all developers of such frameworks and related materials were to fully document their work with manuals and guides in such a manner, reserach projects for graduate students and full researchers alike would take far less time (far outweighing the time required to produce such documentation). This translates in graduate students actually graduating months, if not a year or two, sooner than otherwise and freeing up the funding that would have been wasted on those extra years of education. Also note that the vast majority of our funding comes from taxpayer money, requiring us to police ourselves against the inefficient use of such funds and mismanagement which so frequently characterizes government-run operations.
So here is my contribution; it is likely that others will follow. I hope this small gesture will help provoke a wider and more enlightened view of research conducted no longer by individuals, but by broad team dedicated to a united purpose.How to build an Stntuple Analysis