T Burnett, C Patterson, A Cillis, J Ballet, D Band, J Chiang, S Digel, J Carson, M Hirayama, P Nolan, D Davis, C Shrader, T Stephens, W Focke
No news
JC - Working on adding adhoc energy correction as described at the IRF meeting. Energy distribution functions result in systematically low energy estimates - making correction on average true energy vs. mean energy. Problems with analyzing DC2 data may actually be related to incorrect Aeff estimates
JB - Average true energy for a given measured energy must depend on source spectrum, right?
JC - I'm doing this on an event-by-event basis - know the direction and energy of an event - symmetric in true energy vs measured energy - so can calculate mean true energy vs. measured energy. Problem is that we don't know the true event direction. You can use the reconstructed direction - which is what I did.
JB - I understand what you say about the event direction but that is not the problem- problem is that you do not treat each event separation - just have a statistical thing. Answer depends on the source spectrum.
JC - We can discuss this outside the meeting - can take this off line
SD - Bias is from how the dispersion functions are centered?
JC - The functions are nonsymmetric
AC - Wrote a Python script to make TS contour maps http://www-glast.slac.stanford.edu/sciencetools/meetings/temp/Cillis_TS_contour1.pdf slide 2 - Use Python interface to likelihood analysis; part of the work of the algorithm is to find the correct ranges to use on the parameters being considered; examples on slide 5 of TS contour maps; slide 5 is fast because no free parameters
JB - Did you compare this with what you get from the error matrix that Minuit writes? It should allow making contours like this presumably?
AC- No, not yet
TB - I thought Minuit did contour plots like this itself
SD - Can the script understand general models?
AC - Yes
SD - Put in the repository
DD - Just sort of a demonstrated that can do this - next step is to turn it into a real science tool - something that you run after you've got your best fits to get range on parameters
DB - How fast is it?
AC - About 11 hours for plots on slide 6
No news
MH - Still working on the blind search tool. Related to GUC beta testing, we will improve the documentation to make the examples more realistic - not in terms of parameters but how they use the pulsar database.
SD - With FFTs going must be fairly close to a prototype
MH - Yes, some issues working with the class library but have an almost-working prototype
JC - No news
DB - For the GUC beta test we have been ignoring observation simulation - we are not providing documentation about it, no crash course for it - not pushing it CS - It was a conscious decision to leave that out rather than explain it
CP - Vincent has combined the gtsrcid tutorials - will be posting edited version later today; David's Crash Course is also linked in - on software installation page; DB has provided a number of updates to GRB tutorials and has been incontact with other authors of tutorials with comment; I plan to be current with all changes received to date by COB today - and updates to mirror sites will follow
DB - For GUC, about 16 real testers and a couple of hangers on. Certain things we will not really emphasize in the beta test. Having trouble getting the Python likelihood on Windows and Mac, and as mentioned observation simulation is not ready to be shown to them - interfaces still a bit difficult. In terms of the threads that are out there - in terms of Cicerone, not all tied together. And some threads are fairly DC2-specific. I've asked owners for changes in cases that I have found. In general there's the issue of how to keep threads up to date.
SD - GUC users will be directed to the workbook or through the GSSC site?
DB - Was going to direct them to a begining Web page that links to everything else. I don't think there's a direct link to the Workbook top page. There's links to the threads in the workbook but not the top page - threads = tutorials - but not to the workbook itself. Also links to each of the reference pages, too.
SD - In terms of the documentation are you as far along for GUC as you need to be
DB - Probably about as far along as we are going to get - would like to be further along of course. There are issues in terms of software - Mac version in particular - ROOT libraries, Python libraries - also discovered that in Windows Python plotting with ROOT - many libraries have 0 size. May de-emphasize Python.
CS - Have no choice at this point - the tutorial that I had asked Jim to put together [on Python] will have to be removed
DB - To bring everyone up to date - the beta testers will be using laptops - out of 16 people, 3 have Linux, 7 have Mac, 6 have Windows, +1-2 dual-boot machines
DD - Problems with distribution of tools on Mac - new since DC2 distributions - related to upgrading of compilers
DB - I should mention that at the moment putting tools on Macs with Intel chips was not working as of yesterday. [New issue since DC2]
DD - I was planning to test that a little more this afternoon - Rita has a Macbook.
https://confluence.slac.stanford.edu/display/ST/Science+tool+development+directions
[I didn't transcribe all of this discussion.]
DB - One of the purposes of the beta test is feedback about functionality. Rita and Robin have brought up issue of producing light curves that are not just counts - something with an exposure correction - to see whether a source is staying constant or not - so may be various tools that need to be beefed up or new tools as result of the test
JCar - One topic on my mind for likelihood - a tool that produces an angular power spectrum - angular power spectrum of blazars, clusters of galaxies - a recent paper came out - looking at observable differences - powerful way to look at the data, study relations between source populations
TB - That's interesting - of course, that is a feature of the Healpix software that we have imported from the WMAP people - a student and I are looking at providing that from Healpix representation
JC - Pat, do you want to say something about the stuff you have done in that regard - convolutions on Healpix sphere that I asked you to do?
PN - Some time ago I explored the tools that come with Healpix - looking at convolutions of PSFs in Fourier space -turned out to be fairly straightforward - make a FITS file containing PSF in the right format, then Fourier transform it, multiply by transform of data and then transform back. It didn't require anything but the tools provided
TB - Did you write that up?
PN - I sent e-mail to Jim; I can dredge my archives
SD - The angular power spectra that Jennifer is talking about may not involve convolutions with the PSF and probably relate to distributions of point sources. Not that such a tool would not be worth having, but I don't see it as part of the standard distribution. I was intending the discussion here to be about tools that are to be part of the standard distribution.
DB - This Confluence page may be useful for communicating in greater detail what we want to do
SD - Sounds good
SD - GUC people arrive in 2 weeks?
DB - Yes. Asking GUC people to download software before they come. We are supposed to have an open Starbucks-type wireless system but that may not be in place yet - looking to avoid panics regarding installation - will have helpdesk running - so that they can act as real users in the future.
SD - Next meeting in 2 weeks