Home » Blogs » rjreed's blog

Beamline determination pp500 GeV W triggered events pt 2

Updated on Thu, 2009-08-06 12:50. Originally created by rjreed on 2009-08-03 23:59.

code is at:

/star/u/rjreed/PPV2009/v4

bfc options to generate log files:

Processing bfc.C(1,150,"VtxSeedCalG pp2009a ITTF BEmcChkStat QAalltrigs btofDat Corr3 OSpaceZ2 OGridLeak3D beamLine -VFMinuit  VFPPVnoCTB -dstout evout","/star/data03/daq/2009/Weve/sampler100/st_W_10103033_raw_1180001.daq" ,"st_W_10103033_raw_1180001_1")

 

Last week - Vx,Vy,Ux,Uy values presented from fit which did Vx,Vy and Ux,Uy seperately and used a truncated log-likelihood where xy and z for each track were considered seperately.  There were no uncertainties published.

This week -

I tried a fit of Vx,Ux then Vy,Uy followed by 4 dimensional fit.  Regardless of potential used, this fit appeared to simply walk out of bounds.  I couldn't even get it to converge on sensible values for MC data set for each pair seperately....  I think this is a dead end.

I tried to fit Simplex-Migrad-Simplex-Migrad for each of the three steps.  (Fit Vx,Vy with Ux=0,Uy=0 then fix those values to fit Ux,Uy then use all four values as beginning seed for 4 dimensional fit)  This converged more often when used with the proper log-likelihood distribution.  Simplex fits always converged.  But, it did not always converge upon sensible values.  One thing I decided to do was to restrain the fit.  This does mean that I have to check to make sure fit values aren't right at the edge of the range, but restraining the fit to |Vx|<2, |Vy|<2,|Ux|<0.2, |Uy|<0.2 seemed to give sensible results for MC sets except for E3 (this one has always been finicky ... I've decided to ignore it because none of the data set fits seem to go wrong in quite the same fashion).

I played around with the potential some more.  I tried using a pure gaussian ....  No outlier rejection, but we understand how the uncertainties work.  But, it didn't really work.  The outliers just contributed too much.  So, the lorentzian distribution I was using last week would be even worse as it gives slightly more weight to the outliers than a gaussian.  So I'm still in search of a better potential, but still using truncated log-likelihood.

Set X0 (cm) Y0 (cm) Z0 sig Z x0 (no tilt) y0 (no tilt) ux (fixed xy) uy (fixed xy) x0 y0 ux uy err x err y
E1 0.10 0.20 -20 50 0.10 0.20 0.00000 0.00049 0.10 0.21 -0.00004 0.00057 0.002 -0.011
E2 -0.30 0.40 20 50 -0.31 0.39 -0.00005 -0.00022 -0.31 0.40 -0.00003 -0.00028 0.010 0.003
E3 0.50 1.00 -30 0 0.51 1.01 0.00002 -0.00001 1.10 0.41 0.02000 -0.02000 -0.604 0.591
E4 1.50 -1.00 30 0 1.43 -1.00 0.00000 0.00000 1.43 -1.00 0.00000 0.00000 0.066 -0.003
E5 0.50 -0.40 0 0 0.50 -0.38 -0.01026 -0.01727 0.50 -0.38 -0.01022 -0.01728 -0.002 -0.021
E6 -0.50 0.40 30 0 -0.57 0.48 0.00000 0.00000 -0.57 0.48 0.00000 0.00000 0.070 -0.075

Table 1.  MC set of values with cut-off log-likelihood potential, restrained, fit via the simplex-migrad-simplex-migrad method.  Not all sets converged in final Migrad fit

The question still remains, what is the uncertainty?  The likelihood curve gives an extremely small value, which I do not believe (>0.001 cm)  But, this is what I get when looking at delta Chi^2 = 1 and it matches the returned value on the fit.  Now had the distributions been strictly gaussian, the uncertainty would be the sum of d_i^2/sigma_i^2/N^2.  But this wouldn't be correct as it weighs all the tracks the same, but tracks with d>dmax shouldn't have the same weight.  The result ended up being much larger than I would expect (~0.03-0.07 cm).  So, I added the uncertainties of all tracks with dT^2+ery2<dmax^2 and dZ^2/erz^2 < dmax^2.  The result looks reasonable.

 

                          Tracks with (dT^2+ery^2<dmax^2 && dZ^2+erz^2<dmax^2)  
Set X0 (cm) Y0 (cm) Z0 sig Z x0 (no tilt) y0 (no tilt) ux (fixed xy) uy (fixed xy) x0 y0 ux uy # t sum erry2 sum errz2 unc xy unc z #sx away #s y away
E1 0.10 0.20 -20 50 0.10 0.20 0.00000 0.00049 0.10 0.21 -0.00004 0.00057 536 7.06 5.84 0.013 0.011 0.124957 -0.83178
E2 -0.30 0.40 20 50 -0.31 0.39 -0.00005 -0.00022 -0.31 0.40 -0.00003 -0.00028 551 4.89 6.07 0.009 0.011 1.118995 0.294793
E3 0.50 1.00 -30 0 0.51 1.01 0.00002 -0.00001 1.10 0.41 0.02000 -0.02000 613 6.09 10.13 0.010 0.017 -60.8786 59.50782
E4 1.50 -1.00 30 0 1.43 -1.00 0.00000 0.00000 1.43 -1.00 0.00000 0.00000 504 2.79 4.48 0.006 0.009 11.85905 -0.48648
E5 0.50 -0.40 0 0 0.50 -0.38 -0.01026 -0.01727 0.50 -0.38 -0.01022 -0.01728 712 5.73 8.60 0.008 0.012 -0.24614 -2.62977
E6 -0.50 0.40 30 0 -0.57 0.48 0.00000 0.00000 -0.57 0.48 0.00000 0.00000 641 4.61 9.83 0.007 0.015 9.76791 -10.4632

Table 2. MC set of values with cut-off log-likelihood potential.  Note that sets where ux=uy= 0.00000 appear to be wrong.  These sets still need to be looked at in a slightly more rigorous manner. 

  FIT SEPARATE   FIT TOGETHER            
  Vx Vy Ux Uy Vx Vy Ux Uy # tracks sum (xy) sum (z) sigma xy sigma z
F10383 0.40 -0.01 0.00143 0.00047 0.42 -0.01 0.00162 0.00025 1514 10.51511 10.37218 0.007 0.007
F10398 0.47 0.00 0.00028 -0.00077 0.46 0.00 0.00026 -0.00074 7220 59.78471 62.09663 0.008 0.009
F10399 0.44 0.00 0.00164 0.00025 0.44 0.00 0.00164 0.00025 12333 109.7843 103.0641 0.009 0.008
F10402 0.54 -0.02 0.00019 0.00027 0.54 -0.02 0.00020 0.00027 2155 19.54261 14.78547 0.009 0.007
F10403 0.41 0.00 0.00088 -0.00035 0.41 0.00 0.00088 -0.00033 1695 14.45861 11.35453 0.009 0.007
F10404 0.47 0.00 0.00003 -0.00001 0.47 0.01 0.00011 -0.00006 4854 37.11519 38.76864 0.008 0.008
F10407 0.30 0.01 0.00005 0.00010 0.31 0.02 0.00023 0.00020 4207 36.13407 37.49214 0.009 0.009
F10412 0.42 0.02 0.00118 -0.00007 0.42 0.02 0.00116 -0.00008 17335 149.7633 123.9353 0.009 0.007
F10415 0.41 0.04 0.00087 -0.00041 0.44 0.04 0.00083 -0.00015 9374 87.4472 76.03718 0.009 0.008
F10426 0.54 0.02 0.00033 0.00011 0.54 0.02 0.00029 0.00004 2596 21.86336 19.434 0.008 0.007
F10434 0.41 -0.01 0.00003 -0.00004 0.41 -0.01 0.00003 -0.00004 7476 68.46957 49.08526 0.009 0.007
F10439 0.41 0.04 0.00032 -0.00043 0.41 0.04 0.00033 -0.00042 8924 80.18868 71.95112 0.009 0.008
F10448 0.40 -0.02 0.00011 -0.00001 0.41 -0.01 0.00016 -0.00062 8232 80.14241 62.39852 0.010 0.008
F10449 0.45 -0.01 0.00009 0.00004 0.45 -0.02 0.00077 -0.00024 10564 99.43389 69.05768 0.009 0.007
F10450 0.49 0.02 0.00090 -0.00042 0.47 0.02 0.00105 -0.00033 7755 78.10654 54.204 0.010 0.007
F10454 0.47 0.01 0.00096 0.00044 0.46 0.01 0.00095 0.00040 4499 43.99062 39.77047 0.010 0.009
F10455 0.47 0.00 0.00083 0.00025 0.47 0.00 0.00084 0.00018 7872 72.88313 51.89968 0.009 0.007
F10463 0.44 0.02 0.00000 0.00000 0.44 0.02 0.00000 0.00005 8657 80.35345 61.12146 0.009 0.007
F10464 0.30 0.01 0.00020 0.00195 0.31 0.04 0.00012 0.00207 1678 16.65706 13.02434 0.010 0.008
F10465 0.43 0.00 0.00303 0.00211 0.43 0.00 0.00303 0.00211 5005 55.72243 39.35085 0.011 0.008
F10471 0.41 0.01 0.00001 0.00003 0.41 0.02 0.00007 0.00011 10565 108.475 85.60861 0.010 0.008
F10476 0.41 -0.04 0.00014 -0.00021 0.41 -0.03 0.00002 -0.00029 3481 36.63804 25.48157 0.011 0.007
F10478 0.42 0.06 0.00017 -0.00035 0.44 0.05 0.00020 -0.00063 669 5.736029 6.984221 0.009 0.010
F10482 0.36 0.00 -0.00019 0.00024 0.36 -0.01 0.00013 0.00024 4492 53.04527 41.30754 0.012 0.009
F10486 0.46 0.01 0.00020 0.00000 0.45 0.01 0.00023 0.00002 7472 89.20199 59.26886 0.012 0.008
F10490 0.42 0.02 0.00042 -0.00005 0.43 0.02 0.00045 -0.00010 7903 84.1613 58.95532 0.011 0.007
F10494 0.43 0.02 0.00145 0.00085 0.43 0.02 0.00145 0.00085 6751 94.59211 61.63915 0.014 0.009
F10505 0.40 0.01 0.00116 -0.00046 0.41 0.00 0.00115 -0.00048 7748 93.6367 64.81153 0.012 0.008
F10507 0.40 0.00 0.00153 0.00049 0.41 0.00 0.00169 0.00045 7553 83.91926 56.57406 0.011 0.007
F10508 0.45 0.02 0.00001 0.00003 0.45 0.02 0.00002 0.00004 2382 26.90286 21.44471 0.011 0.009
F10517 0.45 0.04 0.00028 -0.00002 0.44 0.04 0.00053 -0.00019 6152 77.7697 53.99197 0.013 0.009
F10525 0.44 0.02 0.00081 0.00021 0.45 0.01 0.00088 0.00023 7577 82.39373 55.51366 0.011 0.007
F10526 0.42 -0.01 0.00002 -0.00007 0.43 0.00 0.00143 0.00018 4394 58.25066 40.38685 0.013 0.009
F10527 0.47 -0.01 0.00071 -0.00071 0.47 -0.01 0.00072 -0.00071 10994 129.9261 89.73802 0.012 0.008
F10528 0.43 -0.01 0.00000 0.00000 0.43 -0.01 0.00000 0.00000 2706 30.01807 18.82845 0.011 0.007
F10531 0.46 0.02 0.00200 0.00057 0.46 0.02 0.00197 0.00052 10727 124.5245 74.39672 0.012 0.007
F10532 0.43 0.01 0.00009 -0.00005 0.43 0.01 0.00009 -0.00005 8171 95.63088 56.65614 0.012 0.007
F10535 0.40 0.01 0.00025 0.00001 0.40 0.01 0.00025 0.00001 10243 121.7147 82.20712 0.012 0.008
F10536 0.51 0.01 -0.00019 -0.00021 0.50 0.01 -0.00016 -0.00011 3075 33.91912 22.86557 0.011 0.007
Ave 0.43 0.01          

 

          

Table 3: data.  Values in red are quite a bit away from the average, I just want to check them a little more closely.  Files in yellow have no tilt and given what I saw above in MC the data I think these files will also need a little bit of work.

Last thing is how to get from unc in x and z to uncertainty in Ux and Uz.

»