THREAD OF MESSAGES CONCERNING SAFETY OF VETOING ASQ OUTLIERS CORRELATED WITH LARGE EXCURSIONS IN AUXILIARY IFO CHANNELS FIRST NOTE FROM PETER: Date: Tue, 29 Jul 2003 11:05:58 -0700 From: Peter Shawhan To: Gabriela Gonzalez Cc: Keith Riles , John Zweizig , Ed Daw , Peter Fritschel , Brian O'Reilly , Rauha Rahkola , Robert Schofield , Stan Whitcomb , Daniel Sigg , Rana Adhikari Subject: Re: S2 Data Quality: H1 vetos from minute trends Hi Gaby, Can you estimate the "deadtime" introduced by throwing out outliers according to your criteria for the second case? That is, suppose a really big GW burst came along and showed up as a big peak-to-peak excursion in AS_Q. You would look at the auxiliary channels to see whether it is "visible". For what fraction of the run does something seem to be "visible" in auxiliary channels but not in AS_Q? I guess another way to look at this is to say that the veto condition really is only based on the aux. channels; the time should be considered vetoed whether or not you also see something in AS_Q. Well, that's only true for huge events which would be >140 counts peak-to-peak in AS_Q, according to your prescription; smaller GW events would not be vetoed just because something is "visible" in auxiliary channels (but maybe they should be?). Peter REPLY FROM GABY: Date: Sun, 03 Aug 2003 14:43:24 -0500 From: Gabriela Gonzalez To: Peter Shawhan Cc: Keith Riles , John Zweizig , Ed Daw , Peter Fritschel , Brian O'Reilly , Rauha Rahkola , Robert Schofield , Stan Whitcomb , Daniel Sigg , Rana Adhikari Subject: Re: S2 Data Quality: H1 vetos from minute trends Hi, sorry for the delayed answer... the deadtime is minimal, if I had taken ALL 102 outliers it would be 102 minutes/1,040 hrs (0.16%). We could make it even less if we look for the actual glitch, which is typically seconds long rather than a minute long, but that takes more work, it may be better done if it we're desperate or if somebody volunteers... About the issue of vetoing large GWs, I agree that in principle that is possible; however, as in S1, I think it is VERY unlikely (I wish GWs were that easy to see in other channels!). I haven't done the calculations-- hopefully this time we'll have calibrations in the other channels and be able to tell. We also have the hardware injections to test the veto safety, and that we'll do. Gaby. SECOND NOTE FROM PETER: Date: Mon, 04 Aug 2003 17:14:54 -0700 From: Peter Shawhan To: Gabriela Gonzalez Cc: Keith Riles , John Zweizig , Ed Daw , Peter Fritschel , Brian O'Reilly , Rauha Rahkola , Robert Schofield , Stan Whitcomb , Daniel Sigg , Rana Adhikari Subject: Re: S2 Data Quality: H1 vetos from minute trends Hi Gaby, I guess I didn't explain myself sufficiently... I'm not worried (for purposes of this discussion) about a GW showing up in auxiliary channels. I'm only worried about a large GW signal, affecting only AS_Q, which could be vetoed because there happens to be some UNRELATED peak-to-peak excursion "visible" is some auxiliary channel during the same minute. In the case of the 102 AS_Q peak-to-peak outliers, you looked in auxiliary channels and made a judgment about whether something unusual was "visible" during that minute. I was asking what fraction of ALL minutes (regardless of whether there was anything in AS_Q at the time) has something "visible" in one or more auxiliary channels. In other words, for what fraction of the time would the auxiliary-channel info meet your criteria IF THERE WERE an outlier in AS_Q. I am guessing this is still a pretty small fraction of the time, but I don't know what your criteria were. In any case, I think it would be more logical to set a bad-data flag whenever there is a sufficiently large excursion in one of the auxiliary channels, regardless of whether there is anything in AS_Q at the same time. That more accurately represents times when we would want to veto any signal observed during that time (e.g. when doing a simulation to calculate the efficiency of the analysis for some class of sources). Of course, glitchMon with filtering could provide a more precise indicator of big glitches in auxiliary channels, which would be more like a veto than a data quality flag. Peter P.S. I think eliminating segments with multiple AS_Q outliers is fine; it's really the isolated ones I'm worried about. P.P.S. Have you looked at the minute-trend data during hardware injections?