FDS: parameter space searches B. Allen, Y. Itoh, M.A. Papa, X. Siemens AEI , UWM

11
Moving towards a hierarchical search. We now expand the coherent search to inspect a larger parameter space. (At the same time the incoherent stage is being developed and tested). We will pursue two types of searches: short observation time (~1/2 day), no spin-down params, wide band (~300 Hz) centered at ~ 300 Hz, all sky search perhaps longer observation time, 1 spin-down param, small area search (galactic plane/SNRs), small bands. There is a delicate trade-off between sensitivity, observation time (spanned and effective), resolution in parameter space and class of sources that one chooses to target. Calibration info needed to finalize choice SC03 demo. 1000CPUs across the grid for ~ 1 month. Note: different choices could be made in order to produce the FDS: parameter space FDS: parameter space searches searches B. Allen, Y. Itoh, M.A. Papa, X. Siemens AEI , UWM LSC meeting, Hannover, Aug 2003 entire S2 observation time, wide frequency band (200 Hz), in the vicinity of the galactic center. 1000 CPUs across the LSC grid for ~ 1 month: www.lsc-group.phys.uwm.edu/lscdatagrid/details.htm AEI (Merlin, 360 CPUs), Birmingham (Tsunami, 200 CPUs), Caltech (200 CPUs), Cardiff (120 CPUs), ISI (35 CPUs), UWM (Medusa, 300 CPUs)

description

FDS: parameter space searches B. Allen, Y. Itoh, M.A. Papa, X. Siemens AEI , UWM. Moving towards a hierarchical search. We now expand the coherent search to inspect a larger parameter space. (At the same time the incoherent stage is being developed and tested). - PowerPoint PPT Presentation

Transcript of FDS: parameter space searches B. Allen, Y. Itoh, M.A. Papa, X. Siemens AEI , UWM

Page 1: FDS: parameter space searches B. Allen, Y. Itoh, M.A. Papa,  X. Siemens                AEI , UWM

Moving towards a hierarchical search. We now expand the coherent search to inspect a larger parameter space. (At the same time the incoherent stage is being developed and tested).

We will pursue two types of searches:

short observation time (~1/2 day), no spin-down params, wide band (~300 Hz) centered at ~ 300 Hz, all sky search

perhaps longer observation time, 1 spin-down param, small area search (galactic plane/SNRs), small bands. There is a delicate trade-off between sensitivity, observation time (spanned and effective), resolution in parameter space and class of sources that one chooses to target. Calibration info needed to finalize choice SC03 demo. 1000CPUs across the grid for ~ 1 month.

Note: different choices could be made in order to produce the best ULs.

FDS: parameter space searchesFDS: parameter space searches

B. Allen, Y. Itoh, M.A. Papa, X. Siemens AEI , UWM

LSC meeting, Hannover, Aug 2003

entire S2 observation time, wide frequency band (200 Hz), in the vicinity of the galactic center. 1000 CPUs across the LSC grid for ~ 1 month:

www.lsc-group.phys.uwm.edu/lscdatagrid/details.html

AEI (Merlin, 360 CPUs), Birmingham (Tsunami, 200 CPUs),

Caltech (200 CPUs), Cardiff (120 CPUs), ISI (35 CPUs),

UWM (Medusa, 300 CPUs)

Page 2: FDS: parameter space searches B. Allen, Y. Itoh, M.A. Papa,  X. Siemens                AEI , UWM

Modifications wrt the S1 analysisModifications wrt the S1 analysis

• inserted loop to search over different sky locations and spin-down parameters

• introduced more robust Sn estimation technique, based on running median (running median code by S. Mohanty)

LSC meeting, Hannover, Aug 2003

• found the bias correction factor for the expectation value of a running median from an exponential distribution as a function of window size (B. Krishnan)

Page 3: FDS: parameter space searches B. Allen, Y. Itoh, M.A. Papa,  X. Siemens                AEI , UWM

Outlier due to large disturbanceOutlier due to large disturbance

Page 4: FDS: parameter space searches B. Allen, Y. Itoh, M.A. Papa,  X. Siemens                AEI , UWM

Large outliersLarge outliers

The good news: this, like all of the large outliers that we have seen, does not have the F(f0) shape that one would expect from a real signal. The half-height width from a signal is no more than ~ 7 bins wide and the peak is very sharp – no structure like this. So we will implement a test (chi-square test) to discard large outliers based on this principle.

Page 5: FDS: parameter space searches B. Allen, Y. Itoh, M.A. Papa,  X. Siemens                AEI , UWM

Large outliersLarge outliers

• Identify large outlier clusters

• Test whether they can be discarded

f0 max N 2Fmax

For every significant cluster that is identified a line is written to a file (eventually to a DB):

threshold

f0

Page 6: FDS: parameter space searches B. Allen, Y. Itoh, M.A. Papa,  X. Siemens                AEI , UWM

PipelinePipeline

Page 7: FDS: parameter space searches B. Allen, Y. Itoh, M.A. Papa,  X. Siemens                AEI , UWM

Fstat shape test exampleFstat shape test example

Page 8: FDS: parameter space searches B. Allen, Y. Itoh, M.A. Papa,  X. Siemens                AEI , UWM

First results of First results of 22 test test (Y. Itoh)(Y. Itoh)

Page 9: FDS: parameter space searches B. Allen, Y. Itoh, M.A. Papa,  X. Siemens                AEI , UWM

PipelinePipeline

Page 10: FDS: parameter space searches B. Allen, Y. Itoh, M.A. Papa,  X. Siemens                AEI , UWM

2Fmax values in 0.5 Hz bands2Fmax values in 0.5 Hz bands

• 2Fmax value in a 0.5 Hz and searching ~ 15*15 deg around GC, 10h (H1 data)

• after 2 test these most of these values will become smaller

• from each of these values an h095% UL will be derived

consistent with expectations.

h095% ~ few 10-23

Page 11: FDS: parameter space searches B. Allen, Y. Itoh, M.A. Papa,  X. Siemens                AEI , UWM

ScheduleSchedule

• We could not produce complete the analysis due to severe failures of our computing facilities in the past month.

• Expect to have these within the next month – at GWDAW we would like to present methods rather than preliminary results.