REVIEWER LINK FOR H1, EPOCH 1 RESULTS:  http://touro.ligo-la.caltech.edu/~stuver/research/hoft/slope/resultsH1E1.html


Sorry this is so rough...

Here is the spectrogram for the first quarter of the 1st epoch of H1:



Note that this is very similar to the plots that Chad produced but this removes data that is not actually there.  So, end times, like times directly before lock, are not overly exaggerated in the spectrogram.

Then I run SLOPE on the data.  I run within each frequency bin for continuous data segments.  Data segments that are less than 5 samples are excluded (this excludes about 0.3% of the data).  SLOPE determines triggers based on the accidental probability of measuring a given slope in white data.   I ran with a threshold of 0.001 (in the time bins, more on that later).  I then counted how many frequency triggers happened in each time bin.  If there was a significant number of frequency triggers in each time bin, I triggered on that time bin.  Significant was defined to be more than a threshold false rate assuming the data in each frequency bin is white.  That false rate was the threshold I quoted before.  Since there are a different number of analysis windows in the time bins (fixed at the number of frequency bins) compared to the number of windows in the frequency bins (fixed to the time of the continuous data being considered), the frequency bin threshold will need to be adjusted a little.  If you are interested, all of these equations are laid out in my dissertation (it's in the DCC) and I can give you page numbers or write a technical note if you want.  Running the threshold above, 2.66% of the time bins were triggered.

Here is the plot of the spectrogram with the triggered times in black:



Since black lines indicating the triggered times are much thicker than the width of the time bin, I included a close-up to show that it really is only a small portion of the data being triggered on:



These triggers will serve as a smaller set of data to look for poor strain data.  I am almost done collecting the ratio of the measured slope for the trigger to the minimum slope needed to generate the trigger.  Summing these ratios within a time bin gives a rough measurement of the excess energy of the triggered frequency bins.  After all the data has been processed, a simple ranking can be done to see the most significant times identified by SLOPE for data quality.