Prediction histogram does not resemble generated data because of
problem
with event-by-event σ(vpdl).
Changes to generation of event variables
generate Source
generate ct0
generate Pt(B), K from random sampling of histograms
(depends on source)
generate σ(L) from random sampling of histogram
(does not depend on source)
derive L = ct0 Pt(B) / M(src)
Generated Variables
v3.1
v3.2 changes
t0
Pt(B)
Pt(meas)
K
σ(vpdl)
σ(L)
Tag
Tag-smr
VPDL
Lxy
VPDL-smr [w/ σ(vpdl)] (1)
Lxy-smr [w/ &sigma(L)]
Source
Notes:
VPDL smearing was previously done using σ(vpdl)
calculated with a common M(vpdl).
No need for M(vpdl) in generation - remove it in .cpp, .input.
Include functions for coherent calculation of VPDL and Tag,
based on which type of variables are being used in the fits.
Changes to the Likelihood
Need to use L,Pt(meas) in these instead of VPDL(=x).
New: L = f[ L,σ(L),Pt(meas) ]
Prev: L = f[ vpdl,σ(vpdl) ]
remove R-conv only part of PDFevt
Changes to chi2 functions
Have to branch prediction functions for chi2 calculation
off of those for logL calc's because logL functions use σ(L) for
resolution convolution while asymmetry chi2 uses
σ(x).
new prediction function based on v3.1 PDFevt
including convolution over K and VPDL
Use average value for σ(vpdl)
Physics Functions
Continue to use the following inputs to physics functions:
VPDL - calculated from L,Pt(meas) and source
K
Tag
σ(vpdl) - calc from σ(L),Pt(meas) and source
This way the same functions can be used for logL and chi2
fits.
Resolution Functions
Should be able to use same functional form for resolution functions -
just different inputs depending on type of fit.
logL Fits: use L and σ(L)
chi2 Fits: use VPDL and σ(vpdl)
Creating prediction histogram from fit results
logL Fits:
Loop over values of Pt(meas) & L
explicitly include L=0 so that zero-lifetime true VPDL
distrib's are picked up
Choose several values of σ(L) by random sampling over
the histogram
Update prediction arrays for each source at VPDL-bin
corresponding to L,Pt(meas)
using PDFevt( L,Pt(meas),σ(L) ) as a weight
increment bin-by-bin counter each time fill VPDL-bin
Normalize prediction arrays (bin-by-bin) to take into account
the fact that each VPDL bin may have been filled a different
number of times
Fill source-by-source prediction histograms using prediction
arrays
Add all single source predictions together to get total pred
histo
Normalize prediction histograms to data
(globally, based on total pred)