Notes from Science Tools VRVS, January 24, 2007

[These notes are unedited and freely paraphrase the discussion; corrections are welcome.  SD]

T Burnett, J Carson, J Chiang, P Nolan, C Patterson, S Digel, D Band, C Shrader, A Cillis, W Focke, J Peachey, D Davis, J Ballet, T Stephens

Databases

No news

Likelihood

JC - Jean reported a couple of bugs this week but I have not been able to look at them yet

DB - We should get into defining what does TS mean
TB - You mean does TS = 25 really correspond to 5 sigma in the usual gaussian sense?
SD - [said something to the effect that this has been a long-standing question]
TB - Interesting anecdote - CDF collaboration announced discovery of B_s mixing at the "5 sigma" level - and they measured the probability by generating fake experiments - they generated 200 M such. In our case it is quite different - one important difference is wanting to know the fake rate, which has to do how you sample the sky. The probability of a 5 sigma fluctuation being background
JC - Go ahead Pat, I know you have been working with Ramani Pilla
PN - I'm supposed to be working on simple simulations to verify the not-so-simple analytical form that Ramani Pilla has come up with
SD - Simple because fast?
PN - Simple meaning that you understand all of the inputs. So I'm not sure whether we'll see the unexplainable behavoir that Jennifer saw
JC - Have you got to the point of looking at TS distributions - to see whether you get something like Mattox et al. saw?
PN - Not yet
SD - How was the difference again between the TS distributions we get and the chisqr-n type of prediction?
JC - Difference was not shape but overall normalization
SD - Not a simple scale factor, though?
PN - More than half of the results involved a non-detection - negative flux
JC - Sqrt(TS) turned out to be conservative.
SD - That sums up where we stand
DB - From reading Mattox et al.'s paper - used a different threshold of TS at low latitudes and at high - could advise people at what level TS to start believing
JC - The concern is that number will change depending on what your free parameters are in your likelihood fit. I found hardcopy of the distributions that I ran before. The normalization was more like 0.1 or 0.2 instead of like 0.5 that would expect from Mattox et al. results - and the shape was a little bit different. So it didn't really look like a chisq distribution
DB - In 10 days the LAT team will be presenting what the sensitivity of the LAT is and whether it meets the requirements - like how many 10-9 is a 5 sigma detection in one year.
SD - Good question - TS = 25 has been assumed - and we have been relying on it being conservative

GRB tools

DB - No news

Pulsar tools

JP - Working on the A4 tool, and have had the core code working for some time, integrating it in the tool. Discovered an error in the calculation of the trials factor - taking very small numbers to very large powers - need to ensure precision
SD - Applies to gtpsearch in general or just blind searches?
JP - Both - had been evaluating significances for just one trial
SD - Time out of bounds in PulsarSpectrum - and binary pulsar

Observation simulation

No news. 

SD - Richard is again unable to attend to describe his new microquasar source.  Max is continuing to work on binary pulsar updates for PulsarSpectrum and to track down the apparent execution problem that started after v7r6p1 of ScienceTools.

Infrastructure

SD - [I read Richard's note about what support to expect for MacOS in the RM - basically will have Intel Mac computers.]

DB - Isn't there a large installed base of G4 computers?

User Workbook

CP - No news about workbook so far this year - but a new glossary for all our Web pages is in the works
TB - I was accumulating the list for the truth for the 1-year gtobssim run and was trying to understand what some of the parameters are for grbs. Whereas all of the MC source definitions are described very nicely in the page that Jim is responsible for, I didn't see any such a thing for grb simulations -
SD - I see that the link to Pisa has gone away

CP - I'll follow up with Nicola about what we should do to document GRBobsmanager in the Workbook.

Development directions

Likelihood

DB - Modeleditor is something we may not need for another year but in another year we'll really need it - for people preparing GI proposals, doing simulations with the SAE and being able to do that is important - in addition I think that many users are going to want to, when analyzing data, in a sense instead of doing 20 M monte carlos, they will want to do a few simulations of what the region around the source looks like to get a sense of why types of error bars they get on which parameters - so I think we really need this capability.

TS - Regarding pointed observations - what we end up doing there has an impact on how we schedule the LAT - hours on source -
SD - For lost time
JC - Standard consideration, e.g., for RXTE

DB - Isn't it also a consideration of how we define the edge of the FOV
SD - I agree that should be looked at for pointed observations
DB - In pointed mode, the region that you are trying to observe will be close to the axis - the real problem that I see is that the data that will be taken in pointed mode will also be used for analysis of objects that were not targeted - so in a sense the real problem is not for the source that you are looking at in pointed mode but for the sources that are 50 deg or more from the axis - treatment will be affected by earth occurlation

Obs simulation

TS - I would not throw out idea of better simulation for orbit attitude - could make Guisseppes with a better interface - just has taken a very back seat to proposal and operaitons tools

[I didn't get notes for the discussion of the other topics, which was limited owing to limited time before the next VRVS meeting of the morning.]

Wrap-up

[No date for the next meeting was proposed; it will be after the GLAST Symposium, Feb. 5-8.]