Hi,
asked by Daniele I have verified that this effect that I actually see on
the root analysis files is already present on the timestamp from whic we
started: the breco ones.
I've taken a look in /nfs/farm/babar/AWG7/ISL/tmp/TS
and i actually obtain almost the same result comparing the dstar-sb1 with
all the other 'purified (having eliminated all the duplicates among them)'
events.
So on 47755 events dstar-sb1 I see 18021 duplicated events in the
remaining sample (~38%).
This means that it is highly probable that a 'pure' dstar candidate has
some duplicated events somewhere and also means that in this way it is
difficult to evaluate the effect of duplicates without looking at the
purities....
The mes distribution of the candidates in those file is
for the files containing duplicates:
http://www.slac.stanford.edu/~asarti/recoil/OLDmes_d0_sb1.eps
http://www.slac.stanford.edu/~asarti/recoil/OLDmes_dc_sb1.eps
http://www.slac.stanford.edu/~asarti/recoil/OLDmes_dstar_sb1.eps
http://www.slac.stanford.edu/~asarti/recoil/OLDmes_dstar0_sb1.eps
for the new files
http://www.slac.stanford.edu/~asarti/recoil/mes_d0_sb1.eps
http://www.slac.stanford.edu/~asarti/recoil/mes_dc_sb1.eps
http://www.slac.stanford.edu/~asarti/recoil/mes_dstar_sb1.eps
http://www.slac.stanford.edu/~asarti/recoil/mes_dstar0_sb1.eps
In all the various modes the major change is in the signal region.
Cheers,
Alessio
______________________________________________________
Alessio Sarti Now in charge as RQM
Universita' & I.N.F.N. Ferrara
tel +39-0532-781928 Ferrara
roma +39-06-49914338
SLAC +001-650-926-2972
"Vorrei tanto sapere chi e' che va in giro a costruire quadrati
sull'ipotenusa"
On Sat, 6 Jul 2002, Alessio Sarti wrote:
> Hi,
> those are the first results about the killing procedure applied without
> any 'intelligent' algorithm in order to preserve high purity modes.
> All the candidates that are reconstructed in more than one seed or
> superblock are 'killed'.
>
> This not seems the right way to do it. In fact what i see running on
> superblocks 1 for various seeds is:
>
> In input I have
> D0 sb1: Found 4335 entries in tree events
> Dc sb1: Found 2064 entries in tree events
> Dst0 sb1: Found 1667 entries in tree events
> Dst sb1: Found 4501 entries in tree events
>
> And this is the number of duplicates candidadates killed
> D0 sb1 :262 Num. of duplicates
> Dc sb1 :349 Num. of duplicates
> Dst0 sb1 :775 Num. of duplicates
> Dst sb1 :2186 Num. of duplicates
>
> This means that we are throwing away 50% of D* and D*0 candidates.
> Of course this would not happen if the requirement on the purity is
> applied. Given this first result I don't know if it will have any sense to
> quote a systematic doing the fit having killed all the duplicated
> candidates.
>
> In order to check if I'm doing something wrong I've checked by hand (and
> if you want you can do the same test: files are in
> /a/sulky14/AWG12/ISL/killtimestamp
> data are
> ts_files.100X (D0) X=0(sb4) X=2(sb3) X=4(sb2) X=6(sb1)
> ts_files.20X (Dc)
> ts_files.30X (D*)
> ts_files.40X (D*0)
> ) that the number of duplicated candidates in sb1 with respect to all
> other events dumped for different superblock and seeds is ~ 50% of the
> whole sb1 sample for D*. (sb1 has 37002 ev and I found 19392
> duplicates...2000 are still in the same sb and mode: let's say that 2000
> events are duplicated in the same superblock (1) and seed (D*)).
>
> I'm trying to find a simple and smart way to keep the purity information
> in orer to do a more appropriate test.
> I'm still producing the 'free_of_any_duplicate' sample in order to look at
> fit results but i think that the results would not be very helpful.
>
> Questions and comments are welcomed.
> Ciao,
> Alessio
>
> ______________________________________________________
> Alessio Sarti Now in charge as RQM
> Universita' & I.N.F.N. Ferrara
> tel +39-0532-781928 Ferrara
> roma +39-06-49914338
> SLAC +001-650-926-2972
>
> "Vorrei tanto sapere chi e' che va in giro a costruire quadrati
> sull'ipotenusa"
>
>
|