I've created a root file containing a tree with run numbers and fit values (chi2 and ndf) for a constant fit on the gain plots (see my last post, specifically ledPlots.C). 

The root file can be found here: http://www.star.bnl.gov/protected/spin/dilks/led/ledTr.root

The tree includes the following branches:

 - runnum - the full run number

 - day - the day number

 - run - the run number on a particular day

 - nstb,row,col - the cell number

 - chi2 - the chi2 value for the constant fit

 - ndf - the number of degrees of freedom for the constant fit

 - div - chi2/ndf value

 - gcorNom - nominal gain correction

 - ledADC - led ADC

 - ledADCNom - nominal ledADC

 - ledChan - led channel corresponding to cell

The following scripts use the above root file

ledDraw.C -  http://www.star.bnl.gov/protected/spin/dilks/led/ledDraw.C

This script takes a day number and a run number as well as a format (1 for color, 2 for color and text) and draws four 2D histograms. The histograms are then output to pdf files and merged with ghostscript into a file called plots.pdf. The color in the histograms represents the chi2/ndf value for the gain plot constant fits - the closer  the chi2/ndf value is to 1 the more stable the cell. The chi2/ndf values were capped at 30 (i.e. if chi2/ndf>30 then set chi2/ndf equal to 30). This was done so that lower chi2/ndf values could be viewed more clearly on these plots. 

A sample output for run 12095006 is found here: http://www.star.bnl.gov/protected/spin/dilks/led/12095006.pdf

Clearly, the large detectors have more unstable cells than the small detectors... 

ledDrawChanSingle.C - http://www.star.bnl.gov/protected/spin/dilks/led/ledDrawChanSingle.C

This script takes an LED channel number as an argument (see the led map link below). Input channel number 0 as an argument to consider all channels. The script will calculate the average chi2/ndf values for all runs contained in LedTr.root for the specified channel. Two output files are created: chan*.pdf and chan*.txt where the asterisk is the channel number. The pdf file is a colored 2D histogram with the average chi2/ndf values represented as the color and the text file is a table with the columns nstb|row|col|chi2/ndf. 

The average chi2/ndf values were capped at 100 (i.e. if chi2/ndf>100 then set chi2/ndf equal to 100). This was done since some cells had chi2/ndf values that suddenly jumped to ~1000 but were otherwise stable. 

Note that some of the channels will draw blank histograms - this is because there are no cells correlated to these channel numbers.

I have created pdfs for all the channels and merged them into a single file: http://www.star.bnl.gov/protected/spin/dilks/led/chan/all.pdf

The text files can be found in this directory (chan00.txt is all of the channels): http://www.star.bnl.gov/protected/spin/dilks/led/chan/

The script used to automate the creation of all of these files can be found here: http://www.star.bnl.gov/protected/spin/dilks/led/ledDrawChanAll.C

A map of the led channels, created by Yuxi, can be found here: http://www.star.bnl.gov/protected/spin/yuxip/ledmap/MyLEDMap.pdf