I am examining the damage to the lead glass during run 15.  Ultimately
I need to compare with some figure of merit in the physics data, such as
the pion peak positions, widths, or hit rate above some threshold.  This
will come later.  Here is the effect on the LED amplitudes, which overestimates
the effect in the data since it integrates along the entire length of the lead glass
blocks.

I used the following runs:
16077033 Fill 18792  Initial LED amplitudes
16079011 Fill 18787  Integrated Luminosity =   7.411 /pb
16085009 Fill 18815  Integrated Luminosity = 24.634 /pb
16092022 Fill 18##  Integrated Luminosity = 47.038 /pb

(Note: I removed the effect of "radiation hard" lead glass at large eta by
making a cut to remove these cells (led amplitude ration <0.95).  These
can clearly be seen in the first plot:

Here is a plot of the ratio of LED amplitudes for Day 79 (Small Dots= 7.4 1/pb), Day 85 (Open Circles = 24.6 1/pb) and Day 92 (Black Dots = 47 1/pb).

Although the numbers are somewhat rough, we can see that the damage increases with dosage in integrated luminosity and proximity to the beamline, as expected.  In the worst areas, the loss of signal is approximately -1%/[1/pb].

This measures the attentuation for a particular wavelength along the entire length of the lead glass bar.  Of course, this overestimates the problem with loss of gain for real showers.  I have tried another measure: the position and widths of the pion mass peak.  Here is a plot of the ratio between mass peak(width) positions for Day 92/Day80.  I have selected only pions from events with take-all triggers LgBS3/SmBS3 for the appropriate pseudorapidity range. 

And here is the same plot for the ratios of the width of the peaks:

I do not see a correlation with either of these quantities.  I will continue to investigate by
examining the number of hits in each cell over a certain threshold.