Physics

- Efficiencies and Shifts [18-XI.2011]   I redid the shifts with the isolation purity estimates, and dropping correlated cuts (0.75). Slides show performance on leading (di)photon.

- Efficiencies + Smirnov [28-X.2011]   Getting the efficiencies right. Recovering Smirnov. Rel 17 NN on Fudge/Chi^2

- Photon Efficiencies [20-X.2011]   Try to get this stuff out there, so that others trust and use it.

- HCP: Isolation and Rel 17 Smirnov [14-X.2011]   A scattershot of activity.

- Checking Systematics for Nets/Migrate to Rel 17 [6-X.2011]   Check performance of neural nets trained with smirnov Tight, Tight Minus One (var), Uncorrelated 40 (all vars and 3_fu). Look at the N-1 neural nets to identify the most troublesome variables.

- Checking isEM Distributions [18-IX.2011]   Two big pieces: are the (a) distributions and (b) efficiencies correct?

- Status Update and Looking Forward [1-IX.2011]   This is just intended to get the most pressing info that I haven't recorded in the past month or so.

- Smirnov Transforms (and a bit of NN) [1-VIII.2011]   I finished one larger change to the TMVA, namely, moving the isolation requirement of the background to 5-10 GeV. This improved performance somewhat. I then implemented Smirnov transform code. The next step is to apply this to my FlatPhoton skim, and rerun the neural nets.

- Neural Nets and Checks [7-VII.2011]   Work stretching from the first reasonable output, to the first set of checks. The performance on MC/data is great; on data it suddenly dropped off when I fixed the treatment of ΔE.

- Neural Nets [27-VI.2011]   The neural nets are starting to look reasonable.

- Driving Down the Last Problems in IsEM Vars [20-VI.2011]   Hopefully, I understand my isEM at this point; I should be able to plug correct, and correctly shifted variables, into TMVA.

- TMVA Notes [15-VI.2011]   Notes on the first stages of this.

- Isolation Shift MC/Data [18-XI.2011]   Study for HCP and Council. Found a bug in a skim...

- Isolation Distributions in Rel 17, Higher Pileup [4-X.2011]   How do the isolation distributions look?

- IsEM and TMVA [5-VI.2011]   First plots on this.

- First Discussion on isEM [1-VI.2011]   Just chatting and random notes.

- Dump of Hgamgam Composition [20-V.2011]   Here is a summary of things that I have been working on for templates and 2×2D sideband. Probably moving on to IsEM now...

- All Data, Steps Toward Matrix Method [26-IV.2011]   Next steps on the Wisc-induced craziness.

- Fast Fire [23-IV.2011]   All out.

- Broad Sweep on Template Fits [18-IV.2011]   I ran over the potential cuts and looked at the template fits.

- Loose Leading, Track Isolation Efficiency [11-IV.2011]   Look at the effects on the total yield, of using loose cuts; find the track isolation efficiencies of the (ostensibly extremely pure) leading photons with extra cuts.

- Templates: Next Steps [2-IV.2011]   New templates. Rederived electron/photon shifts in MC10. First 'band' approach for estimating background.

- Template Fits Part II [31-III.2011]   Next steps on fitting a template.

- Beginning Template Fits [25-III.2011]   First work with RooFit/Templates.

- Photon Purity First Steps [24-III.2011]   Moving towards photon purity estimates.

- Replicating [22-III.2011]   Work work work on replicating. Rewrote cut flow so that it all happens in the same place. See my SVN. Outputs file of events after cuts (nice for quick plots, afterwards).

- Conversion Plots [22-III.2011]   Divided the invariant mass distributions by the photon conversion flag: unconverted, single track conversion, two track conversions.

- Really, for real, all the official Cuts [2-III.2011]   Basically, I worked to make everything match the official cuts, and to see if I could cinch down to Giovanni Marchiori's numbers -- 80 candidates left in the data after all cuts. I got to 132 with the nominal cuts. They are using rel. 16 with rel 15 energy corrections... is that what you are using, Mike to generate the ntuples? BUUTTTT... a pointer from Marchiori's presentation (attached), page 2:
● reminder: electron contribution undistinguishable from photon => subtract afterwards if non-negligible (determine from control samples - not in this talk)
so I'm not sure how much to sweat this. I also made the correction data v. MC.

- Invariant Mass [18-II.2011]   Three things: (1) looked for a Z-peak in the data/background, (2) switched the order of the isolation cuts to EtCone40 first, (3) reorganized the plotting to show the absolute numbers of events.

- Robust Tight, Track Iso Cut, Code Cleanup [12-II.2011]   I spent most of the week cleaning up code (A LOT OF WORK). I got myself an SVN repository. I switched (thanks, Mike!) to RobustTight. (Because they were very similar in their effect, I simply REPLACED PhotonTight.) I made a Track Isolation Cut at 3 GeV. Reproduced plots.

- First Run with Data [2-II.2011]   I skimmed Mike's full dataset with a cut for two tight photons with Pt > 25 GeV (yay!! it finally worked!!). First run against data. A couple of odd effects -- signal ph_shwr_EtCone40_Corrected FALLS with raising Photon Pt. We are now looking at signal v. ALL backgrounds (incl. irreducible). THIS IS A FIRST PASS. I still want to check things.

- Corrected Background Rejection Method [2-II.2011]   I had been measuring the subleading rejection as the rate for rejecting FAKE photons, instead of rejecting fake SUBLEADING photons for a given event. In other words, the denominator for my rejection was events with a fake famma in the subleading position. I thought about it and realized this is really no good. Corrected, now; the Max(EtCone) then becomes the best rejector, and rejection on the leading photon does nothing.

- Brem Photons. [27-I.2011]   I made the switch to "!(ph_isPhotonFromHardProc_MC || ph_isBrem_MC)", which seems to have 'fixed' things. I am working on figuring out skims. These plots are from the r1834 (full pileup, statistics) sample.

- Switch to EtCone40 [25-I.2011]   This week was a bit of a mess -- oscillating between datasets. Here are plots from ALL OLD datasets using EtCone40 instead of EtCone30. There are still substantial differences from Yuan/Laplace's work.

- New Data: PhotonJet 35 [18-I.2011]   Here, I added to the PhotonJet to the background dataset. I also made plots of the cut on the maximum of the EtCone30 values of the two photons. Note, however, that the cut on the subleading photon remains the most effective.

- Revisiting the Signal Dists; Scale Factors. [17-I.2011]   There were two main problems; the first was to verify the signal distributions/efficiencies, and the second was to implement the appropriate scale factors. The necessary next step, it seems, is to begin using the PhotonJet 35 samples. I will rerun on those, and post again.

- First Steps. [11-I.2011]   My first steps on Hgammagamma. Plots of individual photon isolation distributions, and for photons as an element of an event with basic cuts: PhotonTight, LeadingPhoton Pt > 40 GeV, Subleading > 25 GeV. Efficiency v. rejection for potential cuts. Trends in Eff. v. Rej. (Also, a few battles with the grid.)

- Photon Isolation; Preliminary Chat [6-I.2011]   Preliminary chat with Mike and Brig. Determined which datasets to use and set priorities. Signal Sample Hgamgam; Irred background gamgam15; background, photonjets. Start with EtCone_corrected, and cuts of the form A*x + B.

- ROD Recovery [2-XI.2011]   ROD Recovery Notes (Incomplete)

- Wmunu Last New Plots [26-IV.2010]   Since my last post, I looked at: (1) the eta distribution of fakes, (2) the Et distributions of B quarks, (3) the possibility of using ΔR as a discriminant of correct signal/background, (4) different isolation variables. I presented my work to the group, go feedback, and now I need to write!!

- Background and Egamma Vars [7-IV.2010]   A central question that remained to be addressed was whether we could make an enhanced background sample -- set of jet cuts such that the remaining cuts would not depend on whether the jet was a quark- or gluon-jet. After a bit of wandering, we found that for fake reconstructed photon objects with momenta between around 20 and 50 GeV, simply requiring isolation on the photon object was sufficient that the remaining requirements (PhotonLoose, PhotonTight) did not depend on the leading parton. The multiplication factor was about 10 (unclear if we should be worried about biasing from the JF sample). Along the way, I delved into the IsEM variables and the egamma CSC note.

- Move to D3PDs, Anom. Litt, Fitting ΔR Curves [8-III.2010]   For the past two or so weeks, I read articles and theses and converted everything to run on D3PDs. I both made the D3PDs, and rewrote my local analysis. I was able to cleanly reproduce my results -- if not surprizing, a good confirmation. I tried to start matching a curve for the signal v. background of the ΔR distributions, to use their scalings as a measure of the relative rates of the two processes. This was somewhat less straightforward than expected. Goals looking forward are listed at the bottom.

- Jet Composition and Presentation [14-II.2010]   I gave a presentation at the Penn meeting and got some useful feedback. It looks like we should stick with the 10 (and not 7) GeV cut on the photon. I started looking at the Z + γ background, but the number of missed muons seems too high. I did a first pass on the composition of jets. I think it is important to understand the long-term goals, of where I'm headed.

- Review of Literature 1 [4-II.2010]   I reviewed the articles we had originally discussed; notes and a few questions are listed below.

- Cross Sections and Next Steps with Kin. [1-II.2010]   I heard back from Pavel and updated the plots accordingly. I have started trying to work with the ΔR between the muon and photon for discriminating signal and background, possibly with different cuts for different bands in Et. I decided to standardize my color scheme.

- WmunuTrans [28-I.2010]   I started thinking about what the transverse mass _should_ look like, but hit a few snags. I've attached my first thoughts, hopefully I can get a few pointers. I fixed the transverse (and other) mass calculations to use m² = (Σ Ε)² - (Σ p)². I have posted a few first looks at kinematic variables -- ΔR seems most promising. I heard back from Pavel Staroba -- a promise to answer my MC questions by the end of the week (he was on vacation).

- Wmunu2 [22-I.2010]   Following on my last post, I contacted Pavel Staroba and Borus Kersevan to understand the details of the MC cross-sections, but I have still not heard back. In the meantime, I compared the cut flow for W- + γ to that of W+ + γ (it is nearly the same, as one would expect), and did a transverse mass of the straight W--> μν process, which is cleaner on the RHS, but decidedly not a normal distribution on the LHS. I was curious if the neutrino was just boosted, so I looked at the reconstructed energy of the μ + ν. (I also wrote up this website.)

- Wmunu1 [20-I.2010]   This was my first pass of using isolation on a W + gamma --> mu nu gamma sample. Track isolation with a DeltaR cone of 0.3 and a pt cut of 0.7 yielded rejection of 70%, with 90% efficiency for tight photons. I need to better understand the effective cross sections used in the generation.

- Hydrogen Fine Structure. Derivation of Hydrogen fine structure. Just for my understanding!

- Databases. Manual for using MySQL and Oracle databases, based on a 'migration' of the TRT IntegrationResults database. It is task specific, but should be helpful for a general audience.