I took Guillelmo's histograms (here you have the four separate plots
for Dplus and Dstar)
and performed the usual subtractions: here is the root macro.
These two files [Dplus and Dstar]
show the RS and WS distributions after sideband subtractions (top histograms) and then the
subtraction of the two (bottom left). The sideband scale factors used are 0.5 for the Dplus mode
and 0.1 for the Dstar mode.
These two files [Dplus and Dstar]
show the final subtraction (same as bottom left in the previous ones) on the left, and the same
histogram after requiring the integral to be the same as the one obtained in the moments analysis
(reported here for Dplus and Dstar).
Note that the x axis scale is different for Guillelmo's histograms and for ours.
Here is a summary of what I conclude from the comparison:
- There is a disturbing feature in the D*+ WS signal histogram: a leak of D* faking a D**
This is somewhat puzzling, especially when compared to the sideband wrong sign histogram where
there is no hint of such a contamination. We suspect this hints to an unphysical structure
in the histograms. Note that we don't see the same structure in the corresponding histograms from our data. (See here for details)
- There is no indication of a difference in the mass resolution. In fact, one might argue
that the double peaked structure expected for the narrow states is somewhat hinted in the
defaul analysis histograms and is less visible in Guillelmo's plots
- As expected, due to the differences in the selection variables (Guillelmo is using 2D significnaces, we are using 3D significances), the background in Guillelmo's plots is larger. This is clearly
visible in the high mass tail, where the histograms still roughly average at zero but clearly
have larger error bars.
- The statistical power of the two sets of histograms is comparable. See table below.
A rough estimate of the statistical power of the samples can be obtained comparing the size of
the error bars for the two sets of histograms, once rescaled to the same integral.
In order to somewhat separate the background and the signal contribution, I list both
the size of the error bars in the "excess" region (where the narrow resonances are expected
to show up) and in the "background" region (i.e. the high mass part of the histogram).
here is what I roughly estimate:
| D+ | | D*
|
|---|
| Signal | Background | Signal | Background
|
|---|
| Guillelmo | 6 | 3 | | 5 | 1.5
|
|---|
| Default | 7 | 1 | | 4 | 1
Size of the error bars (+/- events) in the fully subtracted histograms
|
|---|
Alessandro Cerri,
Last modified: Thu Jul 22 17:30:14 CDT 2004