Science Tools Working Group
We met this week.
The current version of ScienceTools remains v8r0p4. v9 should be coming in the next week, before (or as) people are gearing up for obssim2 analyses. v9 will include the Proposed Tool Name Changes and the deletion of the no-longer-needed gtcntsmap. Ideally, v9 will also include some updates to the names of parameters. Dave Davis is working on uniform set of parameter names - uniform across the tools and consistent with common usage among FTOOLS. The parameter name changes would not change the functionality of the tools, of course. Chuck is aware of the need to update the workbook for this and with everything else going on that may take a bit.
Data products: No news.
Databases and related utilities
No news. For SC2 the FT1 data are available from both the GSSC data server (Tom S.) and the Astro Data Server at SLAC (Tony).
Likelihood analysis
No news other than Jim's fix for the calculation of Test Statistic values for binned likelihood analyses, which was in last week's report.
GRB tools
From David Band: "I will be working with James on adding the livetime correction in gtbin when accumulating GBM spectra from their TTE files. Also, we have been talking with Keith Arnaud about PHAII and RSPII files, and how the
PHAII spectra can be linked to RSP matrices and background files."
Pulsar tools
James reports that he and Masa "continue to refactor the application classes for the periodSearch binaries. They consolidated the code that creates the objects used to represent time values, resulting in streamlining of the code in gtpsearch."
Observation simulation
Not much news. Jim explained that the unit test of microQuasar is failing in the RM builds because it needs to be run (with gtobssim) using an FT2 file that I guess is not available. The source itself is apparently working.
User interface and infrastructure (& utilities)
No news.
Source Catalog
Did not meet this week.
Science Tools Working Group
We did not meet this week.
The current version of ScienceTools remains v8r0p4. See what is different in current LATEST1.1878 build.
Data products: No news.
Databases and related utilities
No news. For SC2 we'll have the FT1 data available from both the GSSC data server (Tom S.) and the Astro Data Server at SLAC (Tony).
Likelihood analysis
Jim has implemented a more accurate quick calculation of Likelihood Test Statistic values for binned likelihood in PyLikelihood; the former version overestimated binned TS values by perhaps ~10% (depending on source strength). This brings the binned and unbinned TS values somewhat closer together but a gap still exists.
Unbinned vs. binned TS and fit results have continued to be studied intensively and several presentations on the subject were given at the Catalog VRVS meeting yesterday. In 25 words or less, the binned TS values are systematically larger than unbinned; the binned fits also appear to be somewhat more accurate, except possibly for the faintest sources. Bonus words: The unbinned fluxes are systematically lower than binned for long observations (~1 yr) but not for shorter intervals (few months or less). A few ideas remain not yet explored for characterizing the behavior, basically involving simplifying the system (response functions, scanning, etc.) further.
GRB tools
No news. Jim has the irfInterface package now catching the dgaus8Exception reported last week "to handle cases where the integrator reports insufficently accurate results." The exception, which was being thrown for parts of convolutions in gtrspgen that didn't matter much for the results, was causing gtrspgen to terminate in some circumstances.
Pulsar tools
From James: "James Peachey and Masa Hirayama continued work to rationalize the pulsation search tools gtpsearch and gtpspec. This week they completed the first set of necessary modifications of the TimeRep class family and began altering the applications to use the modifications. The resulting simplifications will make adding barycentering on the fly straightforward and compact by allowing the application classes to share
a common method to perform the operations."
Observation simulation
Not much news. According to the release notes, Richard has modified the microQuasar source to reduce text output.
User interface and infrastructure (& utilities)
From James: "James Peachey created a new test installation for the Science Tools in which he experimented with an interim installation script. This script is not intended for long term use, rather to ease installation and testing of the Science Tools in the GSSC development environment. The script is very simple; it searches through the directory structure to find built binaries and installs them in a top-level bin/ directory, then creates an initialization script and a wrapper which together allow the tools to be run without the user modifying his/her environment significantly. The wrappers behave much the same as those used by the usual Science Tools binary installation, but with the advantage that they do not rely on CMT or on the Release Manager database."
Source Catalog
Met this week; see above.
Science Tools Working Group
We met this week and expect to return to an every-other-week schedule for the foreseeable future.
The current version of ScienceTools remains v8r0p4. This should be considered the 'Service Challenge 2' release unless we come up with another name for Service Challenge 2 (like obssim2 or The Empire Strikes Back).
Data products: Jim has implemented the 2 changes to the FT2 FITS specification that came out of the Pointing and live time history discussion.
Databases and related utilities
No news. It looks like for SC2 we'll have the FT1 data available from both the GSSC data server (Tom S.) and the Astro Data Server at SLAC (Tony). For ASP testing, it also will be ingested (most likely as orbit-sized 'runs') into the Data Catalog at SLAC (also Tony).
Likelihood analysis
Jim reports that he has made gtfindsrc more robust against the rare circumstance that the likelihood surface near maximum has a jump in it that otherwise stalls the amoeba search algorithm.
Work is still continuing on understanding the differences in results (parameter values and TS values) between binned and unbinned likelihood analyses. Contributors include Jim, Vincent Le Biez, Florian Kraft, Benoit, Jean, Ludovic, Toby, and Vincent Lonjou. Space does not permit an adequate summary here. The likelihood Test Statistic values for unbinned analysis are coming out significantly less than for binned analysis, a factor of ~1.4, increasing (or possibly not) with decrease source flux. There's some prejudice to believing the binned TS values, possibly because they are larger. The fluxes reported by unbinned likelihood may (or might not) be slightly biased on the low side. I'm trying to give the impression here that we are getting somewhere but still getting there.
GRB tools
Jim reported a problem with the gtrspgen - actually from Nicola and Nukri - that has to do with getting an exception from the integrator (dgauss8) for the energy dispersion. This is in a part of the code that is checking accuracy and seems to get hung up way out in the tails (e.g., at ~16 GeV when the central energy for a convolution was 160 MeV). Jim thinks that a workaround can be implemented via the irfs package, where the energy dispersions are looked up.
Pulsar tools
James reports that he and Masa are working on consolidating the base classes in the pulsar tools and should have this done by the end of the current build cycle (~1 week from now). This will allow the planned barycenter-on-the-fly feature to be implemented in a relatively straightforward way.
Observation simulation
James reports that Richard ran into a segmentation fault problem with the microquasar source, which was traced to it trying to query for events after the last source gave its last photon. Jim and Toby helped to sort out the resolution.
User interface and infrastructure (& utilities)
James reports that at the GSSC they are working on the HEASARC aspects of the build system for the science tools (for when they get into the business of supporting and delivering the science tools for guest investigators. Larry Brown had made a lot of progress toward automatically getting the science tools built using the HEASARC's hmake - involving parsing CMT requirements files. If we do away with CMT, as seems fairly likely, their Perl scripts presumably will need updating, and some care will need to be taken to minimize the effect of the change from the perspective of getting the tools to build under hmake. Jim and James are going to consult about this.
James has been in communication with the developers of ds9 regarding the possibility of modifying ds9 to make proper region-of-interest selections (that could then be passed back to something like ModelEditor. It isn't an easy problem.
Dave Davis posted proposed name changes for some of the science tools that came out of the discussions at the GSSC-LAT Science Tools Meeting at SLAC, April 2007. He's requesting feedback in the next week. The new names are in general shorter and more uniform; yes, we're keeping 'gt'. For a couple of the tools, 2 possible new names are proposed.
Source Catalog
Did not meet this week.
Science Tools Working Group
We have not met for several weeks now, owing to the GSSC visit to SLAC, the ISOC workshop, and sharing the time slot with the Catalog group. We'll have a Science Tools meeting next week.
The current version of ScienceTools is v8r0p4. Here are the differences from v8r0p3. Among the updates, it looks like Jim made some changes to facilitate builds on Windows and Mac OS-X. The Pulsar and astro packages were updated to address the time-out-of-range problem that has been extensively discussed.
Data products: No news. Tony inquired the other day about whether I knew that the event IDs would be 64-bit quantities.
Databases and related utilities
No news. You might want to look at Tony's presentation and demos from last week's ISOC workshop.
Likelihood analysis
No news. Work is still continuing on understanding the differences in results (parameter values and TS values) between binned and unbinned likelihood analyses. The topic was extensively discussed at the Catalog VRVS meeting yesterday.
GRB tools
No news
Pulsar tools
From James: "James and Masa continued work on refactoring the application-level code for gtpsearch and gtpspec. Specifically, they finished designing an approach to classes to help consolidate the functionality of these tools, (and potentially gtpphase as well). The new code relies on a common infrastructure to determine the time representation based on the contents of the FT1 file."
Mara reports that after this refactoring, they will start working on barycentering-on-the-fly.
Observation simulation
No news
User interface and infrastructure (& utilities)
From James: "James worked on changes to Ape in preparation for its inclusion in the Science Tools. He developed a configure script that handles building Ape with and without readline in a rational way. Also, in response to a complaint from a HEASARC user, he improved the load speed of the parameter file, improving performance for large parameter files.
Source Catalog
Met this week; see above.