Using the yield formula I showed at the last meeting, I can now generate yields for the xs data that have better-contained error bars and are also far more accurate when the signal-to-background ratio of the test sample differs from the sample used to calibrate.

Below is a sample of a yield coming from this algorithm.  The sample luminosity should be roughly 0.5 pb^-1.

the black represents the number of events with geant energy in that bin, the green represents the number of eventsthat survive all the cuts and have tower energy in that bin, and the red represents the final result after correcting the green.

(more current plots of this have labeled axes, etc, but right now the code is in a non-functioning state while I work on it)

I can and have adjusted the composition of the test sample by overall relative weight of the signal or background portion, and also by the weight of individual signal or background bins, and when these adjustments are not too large the agreement between 'true' and reconstructed stays very good.  In more extreme by-bin adjustments (factor of ten increase in signal in some bins, factor of ten suppression in others) the reconstruction begins to systematically go high (though this may just be the particular choices of weighting that I made.

Still unaccounted for in this study is a trigger simulator for the MC, which will add another efficiency to the formula used.

This page may have content added in the near future, so refresh if I start talkign about something that's not here.