This uses the same method as before (counting clusters, e+, e-, mu+, and mu-, taking the best bins).  This time, I’ve added in the charge correlation of leptons and clusters: one cluster with an opposite-signed soft lepton inside goes in the ‘qCorr = -1’ bin, a cluster with two opposite-sign leptons (e.g. a cluster consisting of only an e+ and a mu-) goes in the -2 bin, etc.

I’ve also used Fisher’s method of adding together all of the bins: assign a p-value to each bin, and the total Chi2 is given by:

I’m not quite sure what the best variable to plot for “goodness of this method” is.  Here’s one plot that looks pretty good.  The x axis is number of expected signal events, and the y axis is (expected number that the signal will add to the Fisher Chi2) over (width of the Chi2).  Any point along one of the lines can be reached by choosing a certain number of bins to include.

I’m not quite clear on how the statistics would work here.  If we pick a point with a good expected significance with ~78 events, would we expect to build on that significance once we start looking at kinematic variables?  On the other hand, is this the wrong varable to be plotting?