Our stats have been expanded to include the whole pp500 dataset (thanks Matt), so the next step is to try to fit the ADC distribution in each momentum slice.  At Joe's suggestion I gave up trying to make a Landau fit work and used a Gaussian instead.  Then I plotted the mean value from the gaussian fit vs. momentum and fit it to a line.  Below are the results, first for eta, then for phi.  For eta, the left-hand column is |eta|<.5, the right-hand |eta|>.5.  For both plots, the top row uses the total ADC from a 3-strip cluster; the middle row, total ADC for a 5-strip cluster; and the bottom row, the maximum ADC (all ADC values have relative gains included).  The errors are the errors on the mean from the gaussian fit.

Figure 1: Eta plane linear fits

Figure 2: Phi plane linear fits

The fits for the eta plane are ok (especially for the low-eta bin), but the phi plane fits are pretty bad.  Because some of the data has a vaguely logarithmic look, I also did a log fit ([0]*ln(x-[1])+[2]), which seems to work better for the high-eta bin and for the phi plane (though I should note that I obtained the phi-plane fits only by forcing parameter 1 to be less than 1.75).

Figure 1: Eta plane log fits

Figure 2: Phi plane log fits

So, what next?  Try to figure out a better distribution to fit to, which might improve the phi linear fits?  Work only with the 5-strip clusters, which have ok linear fits in the eta plane and the best linear fit in the phi plane?  Try to incorporate a logarithmic relationship into our overall schema?

Finally, attached is a pdf containing all the fits to slices.