Module 112 was off for 3/4 of fill 10471 resulting in the following problem:

The problem arises because the number of events used to determine xlow and xhigh is too large given that only about ~1/4 of those events contributed to this particular plane.  It can be fixed by using the correct number of events.

Edit 07/02/09: The assumption that it could be fixed by using the correct number of events turns out to be incorrect, as can be seen in this plot:

My guess is this doesn't work because the numbers are just too small.  This module was actually on for only about 16% of the fill.  While the pedestal peaks scale pretty well from this module to others, the rest of the distribution -- the part we are interested in -- is going to be too dependent on statistical fluctuations to scale well.  As a result, the ends of the fit range that I calculate from integrating from the end of the distribution using the parameters I got from the full fill are not going to be all that good, as they will vary a lot depending on statistical fluctuations.  While I could go back and determine a new set of parameters for 16% of the fill, that seems to be overkill for one plane of one module (for whatever reason, the fits in the eta plane, which was off for the exact same amount of time, are much better), so for now I'm just going to leave this as it was.

There are also about ten strips in crate 8 with problems of this variety:

This is due to large pedestal RMS of these strips, for which the 5*RMS cutoff is just not large enough because the RMS is calculated over a limited region.  Options include: discarding strips like 16351 which have very few entries beyond the pedestal and so are effectively dead for our purposes; expanding the cutoff for strips with sufficiently large pedestal RMS; adjusting cuts by hand, though there will probably be too many such strips for this method to be reasonable.