Results
of Running SLOPE for the h(t) Noise Comparison
H1, First Epoch
PRELIMINARY
SLOPE is currently being run over the FFT(h(t))/h(f) ratios to identify
times in the data where this ratio differs significantly from the rest
of the data. This ratio data is computed using 60 second averages
and evenly sampled frequency bandwidth between 100 Hz and 300 Hz.
SLOPE is a redeveloped (from S1) Burst Event Trigger Generator.
Given a specified window (data) length, SLOPE will calculate the best
linear trend to that data. The starting point is then incremented
a sample and the process is repeated until the end of the data is
reached. Triggers are determined based on a probability threshold
that a given measured slope is NOT accidental for white noise
statistics. Therefore, if a threshold of 0.9 is given, then we
expect a false trigger for every 10 triggers.
For this purpose, SLOPE is run in each frequency bin (SLOPE needs white
data to operate on which is the case within the frequency bins because
this data should represent a random fluctuation). The results are
then collected and the excess SLOPE energy in each time bin is
calculated. Excess SLOPE energy is defined to be the sum of the
ratios of the measured slope for a trigger to the minimum slope needed
to pass threshold. The time bins with the largest excess energy
are of the most concern to this investigation.
The SLOPE parameters used in this investigation is: window size = 3 (this give the
first centered derivative calculation) and threshold = 0.99.
For the H1's first calibration epoch, the most significant times are
shown below:
1st and 2nd most significant:
GPS times: 820425815 & 820425695
Closest time identified using the strongest pixel method:
820425755 spectrogram
q
scan

3rd and 4th most significant:
GPS times: 817939576 & 817939456
Closest time identified using the strongest pixel method:
817946173 spectrogram
q
scan

5th and 6th most significant:
GPS times: 821462696 & 821462576
Closest time identified using the strongest pixel method:
821462636 spectrogram
q
scan

Summary:
It should be noticed that while this method does not yield the exact
same results as the strongest pixel method, on a superficial level the
most significant times are similar. This indicates that this
SLOPE investigation is on target for looking for times that differ the
most from the other times. The differences between this method
and the strongest pixel method give different ranks to the most
significant events.
I am currently investigating the times that have excess energy in them
but are not the strongest. I hypothesize that the differences
between the methods will become more apparent since it is not what a
single pixel strength is that is being measured but the contribution of
triggered pixels throughout the time bin. So, while no one pixel
may be that significant by itself, in this SLOPE method the sum of less
significant pixels may bring more subtle h(t) noise issues to light.