Agenda and Minutes, 28 Oct 2010
Agenda: Thurs, 28 Oct 2010
- Continued discussion of S5 H1-H2 paper
Minutes of Oct 28, 2010 Stochastic Review Telecon Attending: Warren, Gregg, Nelson, Robert, Stuart Further review of H1-H2 paper: page 4 III. A Identifying Instrumental and Environmental Contributions - SNR_TS relation to coherence is well understood - can't go lower than 2 without changing noise. SNR_PEM was set to 2 for consistency. - why weeks, months and data set right time periods? Why not days? Too logistically difficult, and weeks seem to capture the behavior. Warren would like to see how one comes to conclusion that days are not going to do a better job. Nelson reminds us that we're calculating all other quantities on weekly timescales. Nelson will generate summary of daily results that were obtained to give better insight. - taking max might overlook nearly significant coherences in multiple channels. Robert will look at vetoing channels and make sure they make sense. Warren is interested in seeing top 10 channels for a representative week to see if there might be multiple correlated noise sources that could be contributing at once. - upconversion and usefulness of bicoherence study should either be removed or explained in more detail. Maybe coherence is enough to capture everything we care about. If not, then we should give more of an explanation of what might have been missed and why we ignored it anyway. page 5 IV. Low Frequency Studies - figure 1 might be a good place to describe correlated noise more - "almost all structure is caused by fans ...", point out specific lines that have well known noise. - Robert feels that so much detail about the calibration line study is not needed. Could add some information about scattering as source of correlations in light. - Also, paragraph beginnning "Finally, let us note..." does not seem to capture what Robert did, in his estimation. A more complete description is maybe in order? - Warren does not believe that Figure 1 really show that the two methods of estimating PEM are complete (as "robust" and "effective" might imply). Rather, it might be interpreted that the PEMs are only finding the narrow band components and that we have no method that's good for broadband correlations, so rather than comforting, this plot might be interpreted as discomfiting because our two methods are not as orthogonal as we had hoped. - Warren notes that the discussion of SNR_PEM values lower than two in the caption of Fig. 1 raises the question "what is the level at which SNR_PEM is primarily statistical fluctuations?" He wants a more quantitative description of why values less than 2 are still meaningful measurements. - Warren wants to know what the size of the bands around the 60 Hz harmonics in which SNR_TS was not computed are. - Warren also wonders if we are missing an opportunity by not calculating SNR_TS near the 60Hz harmonics. Could we learn more about the fall off of the correlations that could let us make more quantitative statements about the correct width of the bands to ignore? For instance, could we fit the fall off?
$Id: 20101028.html,v 1.2 2010/12/09 00:46:59 warren Exp $