Following up on Andy's earlier posting and building on data-folding infrastructure 
he has been developing, Weigang Liu has tried folding the same three channels Andy examined, 
but folding daily and monthly over most of the O1 data set
(ignoring January data to avoid confusion from calibration / post-O1 studies).
Weigang sees similar behavior in each channel to what Andy found in a half-hour sample, 
but with some variations during the course of the run.

In brief, Weigang uses a Matlab script to read in full PEM data or observing-mode DARM-related channels,
folds it upon itself in 4-second intervals, computes an FFT of the folded data, excises unwanted
bands, then inverts the FFT to get a band-passed/low-passed/high=passed time series. Discontinuities
at starts of science segments are handled with a smooth ramping, and the 4-second intervals are
shifted by a half-second to avoid artifacts on integer-second boundaries due to the data processing.
All 4-second intervals are synchronized to be commensurate over gaps in data.

Attached below and linked here are summary plots for Sept-Dec 2015 for two magnetometer channels (40-Hz low-pass):
H1:PEM-EY_MAG_EBAY_SUSRACK_QUAD_SUM_DQ
H1:PEM-EY_MAG_EBAY_SEISRACK_QUAD_SUM_DQ
and a summary plot for Sept-Dec 2015 for the DARM-related suspensions channel (10-50 Hz band-pass):
H1:SUS-ETMY_L3_MASTER_OUT_LL_DQ

Corresponding monthly and daily plots can be found by drilling down at these links:
H1:PEM-EY_MAG_EBAY_SUSRACK_QUAD_SUM_DQ
H1:PEM-EY_MAG_EBAY_SEISRACK_QUAD_SUM_DQ
H1:SUS-ETMY_L3_MASTER_OUT_LL_DQ