The set of fits over all pT bins plus all summary plots are placed into a pdf for each combination of options.  These provide the full amount of information from the study of fitting various background shapes.

Options considered

Choice of algorithm:

• IU
• Morhac
• TSP

Choice of whether to subtract the combinatorial background estimate

• subtract (sub)
• do not subtract (noSub)

Choice of whether to smooth the histogram

• no smoothing
• smooth w/ 10 iterations (smth)

Functional shape:

• Gaussian plus skewed Gaussian (GG)
• Gaussian plus 4th order polynomial over large range (PG)
• Gaussian plus linear background over small range (SP)  [Note: I forgot why SP seemed like a logical name, and now it's too late to change for today.]

Description of Plots in the PDFs

• In case of subtracting the background, the first page shows the data and the normalized combinatorial background estimate.  One plot per pT bin.
• Next two pages have
  • left panel: the histrogram which is fit, plus the fit function (black), the signal part of the fit function (blue) and the background portion of the fit function (magenta)
  • right panel: residual of the histrogram minus the fit.  Errors unchanged.  Fit to a 2nd order polynomial shown in green (as an estimate of how "zero" the residual is)
• Final page: various summary statistics, plotted vs per pT bin.  Note: for the SP case, the small and large range is the same, while for the other fit cases, the "small" is over the same range as the range of the residual plot, and the "large" range can be seen from the range of the fit function in the left panel plots.

File Location

Due to the large number of files, they have been placed on the protected site.  http://www.star.bnl.gov/protected/spin/sgliske/pi0/2012-March/.  The names are based on the above listed options.

Conclusions

• Comparing smoothed vs. non-smoothed gives some indication of how much of the chi^2 is due to random fluctuations and how much is due to using the wrong shape of the fit function.
• The 'SP' fit for the Morhac and TSP algorithms yield very good chi^2.
• In all cases, it is prefereable to not subtract the estimate of the combinatorial background.