Changes To The Event Selection
To improve modeling in the PreTag Control Region, several selections have been modified since the summer 2011 search. The effect on expected sensitivity of these modified selections is ~5%. The selections are compared below.
| Selection | Summer 2011 | Current |
|---|---|---|
| Minimum Lead-Jet ET | 25 GeV | 25 GeV |
| Minimum Subleading-Jet ET | 15 GeV | 25 GeV |
| Number of Jets | at least 2 | 2 or 3, in separate sub-channels |
| Minimum Dijet Mass | no requirement | 25 GeV/c2 |
To improve the efficiency with which we identify b-jets, all b-tagging is currently done with the new HOBIT multivariate tagging algorithm. A comparison of the current and previous b-identification categories is shown below.
| Selection | Summer 2011 | Current |
|---|---|---|
| Double Tight Tag Category | Two Tight Secondary Vertex Tags | Two Tight HOBIT Tags |
| Intermediate Double Tag Category | One Loose Secondary Vertex Tag + One Loose Jet Probability Tag | One Tight and One HOBIT Tag |
| Single Tag Category | One Tight Secondary Vertex Tag | One Tight HOBIT Tag |
| Double Loose Tag Category | No Category | Two Loose HOBIT Tags |
Due to the modified selections only 41% of the summer 2011 analysis sample is retained in the current search. Additionally, the combination of increased acceptance to b-jets and the increased size of the analyzed data sample means that 53% of candidates in the current search are new and were not present in the summer 2011 analysis. The following figures show the most signal-like candidates in the electron and muon channels, as ranked by the ratio of signal-to-background assuming a 120 GeV Higgs Signal. As can be seen many of the high S/B events in the current search were not present in the summer 2011 analysis (the horizontal axis labels show event identification numbers) :
The impact of one or two very signal like events on the observed limits is quite strong as can be seen for the electron channels below :
Changes To The Event Discriminants
All discriminants used in the current analysis have been re-optimized since the summer 2011 versions. In addition, the structure of the final discriminants has been modified with the addition of a new diboson expert network, and the replacement of a jet flavor separator with a combined Z+cc and Z+lf expert network. Due to these changes the significance of a given event changes between the two versions of the analysis. Below we show the the location (in the final discriminant) of the two highest S/B events shared between the current and summer versions of the electron search.
Changes To Expected and Observed Upper Limits
Due to the modified search technique expected sensitivity to a Higgs signal is significantly improved compared to the summer 2011 search.
Similarly the observed limits have changed with respect to the expected limits.
| Combined Limits Winter 2012 | Combined Limits Summer 2011 |
|---|---|
| high res. image | high res. image |
| ZH→e+e-bb Limits Winter 2012 | ZH→e+e-bb Limits Summer 2011 |
| high res. image | high res. image |
| ZH→μ+μ-bb Limits Winter 2012 | ZH→μ+μ-bb Limits Summer 2011 |
| high res. image | high res. image |
