Science Tools Working Group
We had a brief meeting this week (attendees: D Smith, J Ballet, J Chiang, S Digel, T Burnett, C Patterson, D Band, E Winter, T Stephens, M Hirayama, D Sanchez, J Cohen-Tanugi, J McEnery); we will not meet next week.
As of Wednesday, the current version of the Science Tools is v9r3p4. Here are the differences from v9r3p2. The new version has important fixes in the astro (evaluation whether GLAST is in the SAA) and Likelihood (see below) packages.
Data products: We discussed getting the current definitions of the data products in the File Format Document that accompanies the Interface Control Document with the GSSC that needs to be in place by the time of the Flight Operations Review. Most of our discussion was about the format of LS-003 (livetime cubes). David's draft specification needed to have a number of header keywords that are not currently included. You might also want to see Jim's note to scisoftlist, if you missed it.
Databases and related utilities
No news
Likelihood analysis
Jim described improving the accuracy of the calculation of the diffuse responses and the marked effect on TS values for unbinned likelihood - he gave a detailed description in the Catalog meeting yesterday. This work resolves the long-recognized discrepancies between binned and unbinned TS values.
Jim has also added warning messages if a source model includes sources well outside the ROI of the current analysis.
GRB tools
No news.
Pulsar tools
David Smith described the discussions with the GSSC last week regarding potential enhancements to the pulsar tools. David and Dave Thompson at least were among the participants. Bordeaux will be contributing to implementing some of the requested updates.
Masa had microphone problems in the EVO meeting.
Observation simulation
No news.
User interface and infrastructure (& utilities)
Eric Winter (who also had microphone problems) is circulating a proposal to have Likelihood and obsSim XML model files be self identifying, using the "xmlns" attribute in the top-level element in the files. The formats are enough different that a program could figure out which is which but having an identifier at the top would make the job easier. This relates to Eric's work on an enhanced version of ModelEditor that could translate (where possible) between Likelihood and obsSim models.
Source Catalog
Met again this week. Jean presented comparisons of the performance of the source detection algorithms applied to the obssim2 data. Toby's pointfind currently has the best overall performance - see the plots that Jean posted. Jim also presented the results described above on evaluating TS in unbinned likelihood analysis.