Science Tools Working Group
Met this week - 9 attendees.
The current version of ScienceTools is v9. This has the fixes described in last week's report for v8r2 and also implements the tool name changes. Chuck (as probably has or will be mentioned during this meeting) has a new version of the Science Tools section of the User Workbook waiting in the wings, which with some verification should be online by early next week, at which time v9 will be advertised to subscribers to the Service Challenge mailing list. As of Wednesday evening Tom S. was ingesting the reprocessed obssim2 data into the GSSC server.
Regarding the GLAST Users Group meeting reported on two weeks ago, the GUG now plans its next 'beta test' of the Science Tools for September.
Data products: No news. Analia submitted a JIRA issue (DATAPROD-3) about problems (violations of FITS standards) found by fverify for some of the FITS data products. Jim has implemented fixes, or workarounds for an artifact that tip introduces, that fix most of the problems that she reported.
Databases and related utilities
Tony gave a great presentation on the Data Server yesterday at the Service Challenge Steering meeting.
Likelihood analysis
No news but see utilities and Catalog news below.
GRB tools
No LAT-specific news. David reports that gtbin needs a small update to handle detector-specific deadtime in calculating 'exposures' (accumulated live times) for GBM data.
Pulsar tools
Masa reports that he and James are reaching the point with refactoring gtpsearch that they can start propagating the reorganization through the rest of the pulsar tools. They have proposed additional columns in the pulsar ephemeris file for handling the timing parameters of binary pulsars; the ingest system for ephemerides at the GSSC is still to be developed but is on the not-too-distant horizon.
Observation simulation
No news
User interface and infrastructure (& utilities)
Jim has modified gtmktime to not change TSTART/TSTOP values in FT1 files. As reported last weeek he has optimized the tool so that it is execution time does not scale as N^2 (number of time intervals) but as a much slower function of N.
See last week's report for an important note about gtselect.
Source Catalog
Did not meet this week.
Science Tools Working Group
Did not meet this week.
The current version of ScienceTools is v8r2. This has the fix to the Likelihood bug in handling the HANDOFF response functions that was described last week. It also fixes the EVENT_CLASS numbering of the front and back HANDOFF response functions to be 0 and 1, respectively, as they are for other response functions.
This means that v8r2 cannot be used for likelihood analysis with any already-generated gtobssim data set that you have using the HANDOFF response functions. This includes the obssim2 data set. Nicola has reprocessed it - fixing the EVENT_CLASS settings and regenerating the diffuse responses. We have learned of a small bug in how the TSTART/TSTOP keywords are set in some of the daily files in the obssim2 dataset. The bug is of no consequence (that I'm aware of) for any Science Tools analysis but would at least have caused the files to fail the ingest checks at the GSSC if Tom had not already had to disable them to accept the daily files that we sent him. So we are considering what to do, and if you want to know the details I'll tell you. We'll figure out what to do soon, so we can get the reprocessed obssim2 data into the servers.
Today at Harvard, David Band is describing the analysis tools at a Getting Involved with GLAST workshop.
Regarding the GLAST Users Group meeting reported on last week, they GUG plans to have another 'beta test' of the Science Tools, sometime September-November.
Data products: No news
Databases and related utilities
No news.
Likelihood analysis
No news but see utilities and Catalog news below.
GRB tools
James reports fixing a problem with the unit test in the rspgen package
Pulsar tools
From James: "James and Masa continued refactoring gtpsearch. During this period, they concentrated on rationalizing the application code with an eventual view of handling barycentering on-the-fly. They continued breaking the
application into pieces using helper methods in the application class. These helper methods are streamlining the flow of the main gtpsearch application, and in the future will allow gtpspec and other pulsar tools to take advantage of the periodSearch application classes.
James also reports fixing build problems in the periodSearch package.
Observation simulation
No news
User interface and infrastructure (& utilities)
From Jim: "gtmktime [future name for gtmaketime] has been optimized and showed a factor of 50 speedup for a 10^7 s gtobssim simulation with standard orbit and rocking. gtmktime does not modify TSTART or TSTOP or other date keywords.
gtselect can now accept a list of FT1 files. Non-GTI keywords are compared for consistency, but GTIs are not examined, so one can have overlapping GTIs (and potentially multiply-included events); and the GTIs are merged without discretion. gtselect ensures that the TSTART, TSTOP, and date-related keywords are synched up to the user's
selections in all HDUs."
The above modifications mean that the GTI bounds will not necessarily correspond to the TSTART & TSTOP values.
Source Catalog
Met this week. Jean reported that a preliminary catalog for obssim2 is ready - ancillary information like light curves and fluxes in different bands will come later but the numbers of sources should not change. He has provided it to Tom for posting at the GSSC and will make an announcement in the SC Users Forum when it is ready. the catalog analysis relied on binned likelihood and so was not affected by the swap of front and back HANDOFF response functions.
Other topics included presentations on source localization (by Toby and Vincent), an update on MR_FILTER's performance on 'test patterns' (Ludovic) and some discussion about the contents of the obssim2 source list.
Science Tools Working Group
We met this week.
The current version of ScienceTools is v8r1, but it is not quite the same v8r1 as last week. Owing to work that he already had in progress to implement the Proposed Tool Name Changes, when Jim fixed the bug in how Likelihood handles the HANDOFF response functions (see below), the only workable option for making a release with that bug fix but not the tool name changes was to move a tag or two and re-release v8r1. If you had installed it before June 13, you should remove and reinstall it.
As Jim proposed and was discussed at the Service Challenge Steering meeting on Wednesday, next week he will release v8r1p1 which will fix the reversal of front and back designations (so EVENT_CLASS = 0 for front and 1 for back). This will be synchronized with a release of a reprocessed obssim2 data set. v9, with the tool name changes and probably the Proposed Parameter Name Changes would follow.
Updated Pass 4 response functions (of which HANDOFF is the current example) are anticipated around the time that the next Service Challenge data set (3-hour simulations via Gleam) will be released. The current HANDOFF response functions will not disappear from Science Tools at that time, so the obssim2 data will remain analyzable. I'd guess that the new Pass 4 response functions (name TBD) will first appear in something like v9r1.
Regarding the GLAST Users Group meeting reported on last week, they GUG plans to have another 'beta test' of the Science Tools, sometime September-November.
Data products: No news.
Databases and related utilities
No news. Tom S. reports about 80 queries for obssim2 data have been served by the GSSC so far - this exludes queries that Tom has made to test the server. Some queries failed on June 1, but it was not immediately clear why.
Likelihood analysis
Other than Jim's finding and fixing the handoff response handling bug and apparent Likelihood bias the news is that studies of likelihood analysis - TS values and parameter biases from unbinned and binned likelihood continue. Florien & Vincent have been looking at the accuracy of the likelihood calculation in the unbinned case by comparing the value of log(Likelihood) for a 360 day-long simulation (simple point source + diffuse background) with the log(Likelihood) derived by evaluating the likelihood function for 10-day subintervals of the data set and adding the results. The difference was about 1%, and this translated to measurable differences in TS. Not clear yet whether the magnitude of the differences is a concern (about 80 out of TS = 4000 for the relatively bright point source considered).
GRB tools
No news
Pulsar tools
Masa reports that he and James continue to work on sorting out the code of gtpsearch for barycentering-on-the-fly. He and James also checked that the proposed parameter name changes do not affect the parameters used by the pulsar tools.
Observation simulation
Max R. (who submitted his Ph.D. thesis today) has checked in a fix for the initialization problems encountered with PulsarSpectrum in the recent attempts that Tom G. has made for a SC simulation via Gleam. According to Jim the problem was an infinite loop (non-convergence) in barycenter arrival time decorrections for some time intervals.
Toby reports that he has fixed the pointed observation mode of gtorbsim, at Dave D.'s request. So now it can generate pointed observations; keep in mind that it does not perform earth avoidance slews.
The GSSC is close - really - to having a Science Tools version of the orbit/attitude simulator written by Guiseppe R. This should supercede gtorbsim
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 is v8r1. Here are the differences from v8r0p4. This includes a number of mostly small updates from the previous release - some of which are mentioned below. This will be the last incremental release before v9, which as reported last week will include the Proposed Tool Name Changes and the deletion of the no-longer-needed gtcntsmap.
Ideally, v9 will also include updates to the names of parameters. Dave Davis has compiled a list of recommended changes - to a more uniform set across the tools and overall more consistent with general usage for FTOOLS for other missions in the HEASARC archive. These are bigger changes in terms of the code than the tool names. The proposed changes are now open for a (brief) comment period.
Changing names of tools and parameters will be inconvenient to existing users, and there's never a good time. Sooner is the least worst option (and we don't expect to do this again). I hope to figure out today how realistic going to v9 will be in the near term with the new parameter names incorporated. If we cannot get these changes in soon, then we'll need to hold off while the obssim2 data set (and possibly its successors) enjoys its heyday.
The GLAST Users Group (formerly GUC) met this week at Goddard; some of the presentations that they were shown are available online. Chris Shrader described the status of the GSSC, including the Science Tools. He reports that the GUG agreed that the first GSSC distribution of the science tools (to the world at large) should be at Launch + LEO + 6 months.
Data products: No news.
Databases and related utilities
No news.
Likelihood analysis
From Jim: "Responding to Jean's problems with analyzing front-only events using handoff response, I added the HANDOFF_front and HANDOFF_back combinations to those available from the IRFs loading software. To use these one would do, e.g., gtdiffrsp rspfunc=HANDOFF_front" [gtdiffrsp being the v9 name of gtdiffresp]
GRB tools
No news
Pulsar tools
Masa reports that he and James have done some bug fixing and continue to work on sorting out the code of gtpsearch for barycentering-on-the-fly.
Marcus Ziegler is starting to make a Science Tools version of the pulsar search with time differences algorithm. See Atwood et al. (2006).
Observation simulation
No news.
User interface and infrastructure (& utilities)
No news.
Source Catalog
Met this week. The agenda included a variety of topics. Some further studies of likelihood analysis (including confidence regions reported by gtfindsrc) were presented, along with investigations of source detection in the Catalog pipeline and with the new 'test pattern' data sets.