From vmandic@ligo.caltech.edu Thu Jun 16 16:37:52 2005
Date: Thu, 16 Jun 2005 13:37:26 -0700 (PDT)
From: Vuk Mandic <vmandic@ligo.caltech.edu>
Reply-To: Vuk Mandic <mandic_v@ligo.caltech.edu>
To: Joseph D. Romano <Joseph.Romano@astro.cf.ac.uk>
Cc: Keith Riles <kriles@umich.edu>, Vuk Mandic <vmandic@ligo.caltech.edu>, Peter Fritschel <fritschel_p@ligo.mit.edu>, Gabriela Gonzalez <gonzalez@lsu.edu>
Subject: Re: S3 intervals excluded from stochastic analysis

Hi,
the attached are the bad GPS times for the H1L1 analysis and H2L1 analysis
- each time is actually the start-time of the 60sec segment. Most of the
bad times are determined by comparing the theoretical sigma (on our
cross-correlation estimator) calculated using the 60-sec segment in
question, with the sigma calculated using the two neighboring 60-sec
segments. We require them to be within 20%.

We could not run this code on a single IFO - the parameter we use is based
on cross-correlation. But we could make modifications and do something
similar using the PSD of a single IFO. It would take a little work,
though. Thanks,
            Vuk

P.S. The 20% statement is roughly correct for both pairs.

On 16 Jun 2005, Joseph D. Romano wrote:

> hi vuk,
> 
> please take a look at keith's e-mail below.  are there files with the
> bad GPS times from the sigma outlier cuts?  if so, could you please send
> them to keith.
> 
> thanks.
> 
> ---joe
> 
> On Wed, 2005-06-15 at 15:42, Keith Riles wrote:
> > Peter and Joe,
> > 
> > Following up on a suggestion from Gaby, I'm curious if you
> > have kept a list of the 60-second single-IFO intervals that
> > were excluded from the S3 stochastic analysis. It might be
> > useful to other analysis groups to know which intervals are
> > characterized by non-stationary noise, as seen by the stochastic
> > analysis. I have in mind adding a new S3DQ flag with a
> > name like NONSTATIONARY_PSD_STOCH. According to the S3 stochastic
> > paper, about 20% of the data is excluded. Does that statement
> > apply to both H1-L1 and H2-L1 coincidence intervals? Is L1 the
> > usual culprit? You probably have marked only intervals for which
> > you have double-coincidence data. Would it also be possible to
> > run your code on single-IFO data, applying the same criteria
> > on PSD consistency? (If not, the double-coincidence info is still
> > quite useful - I would just need to state carefully on the S3DQ
> > page which samples were examined.)
> > 
> > thanks very much,
> > Keith
> > 
> 
> 

    [ Part 2, ""  Text/PLAIN (Name: ]
    [ "badGPSTimes_H1L1_hann_newcalib.dat")  1,366 lines. ]
    [ Unable to print this part. ]


    [ Part 3, ""  Text/PLAIN (Name: ]
    [ "badGPSTimes_H2L1_hann_newcalib.dat")  1,141 lines. ]
    [ Unable to print this part. ]