Fitter Notes

To Do

• look for straw crossings
• get better starting points, especially for theta
• continue reformat of fitter
• put in all trajectories during fit
• charge division
• plot the residuals

Plotting the residuals

two approaches:
1. turn hddm file into ntuple
2. read hddm file and fill histograms
option 1 more flexible
option 1 could be enhanced to automatically generate program to do conversion
option 2 quick and dirty
needs stand-alone root binary
can be developed in root-examples
option 1 would require same binary

Make the example binary, simple histogram: done in CVS, my, root_example/create_hist.cxx. Makefile there builds it.

Made program to plot residuals: SVN, Hall D, home/marki/gluex/plot_resids. Here is a fit to a distribution. Same events as described in subsequent section.

fit to residual distribution for 1 GeV/c pions

Scale is in centimeters (from DCDCTrackHit.h and swim.tex).

prob vs. prob

Made plot of “residual” chi-squared probability vs. “parameter” chi-squared probability.

residual chi-squared prob vs parameter prob

Done with default hdgeant physics settings, 1 GeV pi+’s, generated uniformly in solid angle.

pt not ptot

Change fitter to use transverse momentum rather than total momentum in fit. Should converge a bit better. SVN log message:

r4413 | marki | 2008-10-28 14:29:03 -0400 (Tue, 28 Oct 2008) | 14 lines

Merge pinv to ptinv change from branch to trunk. Command was

svn merge -r4411:4412 $HDSVN/branches/home/marki/gluex/fitter-ptinv

M prob.C
M MyTrajectoryBfield.h
M DTrackLSFitter.cc
M MyProcessor.cc
M DTrackLSFitter.h
M fitter.C
M MyTrajectoryBfield.cc
M swim_test.cc
M bad.C