Coincidences between candidates

In order to establish the probability of detection of a real gravitational wave, after finding the candidate signals with the F-statistic candidate signal search, the pipeline searches for coincidences between candidates found in different time frames.

The coincidences code is available at github. To get it, run git clone https://github.com/mbejger/polgraw-allsky.git.

Prerequisites

The code is written in standard C. The only dependency is GNU Scientific Library (GSL), used to manipulate the Fisher matrix (calculate the eigenvectors and eigenvalues). GNU struct dirent objects are used to read the directories.

The idea behind coincidences

After finding the candidate signals in different time frames (search), we want to confirm the existence of signals with the same parameters along the span of time segments. to further perform a validation search for high-coincidence, or otherwise interesting candidates (the followup, currently under construction). To do this, the candidates from a list of trigger files (time frames) are read, and for each trigger file

• a candidate is transformed to a well-defined time moment (frequency shifted to a reference frame),
• translated into linear (integer) coordinates,
• duplicates within each frame are removed,
• list of unique candidates from all the frames is created and sorted,
• duplicates are counted - these are the coincidences.

TODO: describe cell shifts (16 different shifts in total: 0101, 0110, 0010 etc. in f, s, d, a directions) and scaling of cells (used to define the linear parameters for a given cell to subsequently compare the candidate values)

Compilation

Run make coincidences; resulting binary is called coincidences. Modify the Makefile to fit your system.

Full list of switches

Type

% ./coincidences --help 

to obtain the following description:

Also:

Example

Using the software injection added to 2-day Gaussian noise data segments (see minimal example of the pipeline):

% for s in {0..1}{0..1}{0..1}{0..1}; do ./coincidences -data ../../search/network/src-cpu -output . -shift $s -scalef 4 -scales 4 -scaled 4 -scalea 4 -refr 4 -dt 2 -trigname 1234_2 -refloc ../../testdata/2d_0.25/004 -nod 2 -fpo 308.859375 -snrcutoff 5; done 2>> summary

This assumes that for the band $bbbb=1234$ and the sampling time $dt=2\ \mathrm{s}$ the band frequency $fpo=308.859375\ \mathrm{Hz}$, because

The reference grid file grid.bin, corresponding to the reference frame 004 (-refr 4) is located at the -refloc location. Signal-to-noise ratio cutoff is set to 5 (-snrcutoff 5). Some output is directed to stdin. For example, the output for shift 0000 is

Number of days is 2
The SNR threshold cutoff is 5.000000000000, corresponding to F-statistic value of 14.500000000000
Reading the reference grid.bin at ../../testdata/2d_0.25/004
fftpad from the grid file: 1
Settings dt: 2.000000, oms: 3881.241374
Reference frame number: 4
Cell shifts  in f, s, d, a directions: 0 0 0 0 
Cell scaling in f, s, d, a directions: 4 4 4 4 
Reading triggers... Frame 5: 1966/2040
Reading triggers... Frame 1: 2409/2483
Reading triggers... Frame 4: 2176/2384
Reading triggers... Frame 3: 2132/2247
Reading triggers... Frame 8: 2372/2408
Reading triggers... Frame 2: 2197/2249
Reading triggers... Frame 6: 2225/2305
Reading triggers... Frame 7: 2175/2226
Total number of candidates from all frames: 17652

The highest multiplicity coincidence for each shift is streamed to the summary file via the stderr (2>> summary). From this list of 16 lines, the highest multiplicity coincidence with the highest signal-to-noise is selected (sort -gk5 -gk10 summary | tail -1):

1234_2 1111 308.859375     8     5  9.95663703e-01 -1.10830358e-09 -1.12585347e-01 1.97463002e+00 1.246469e+01 5 2040 1987 1 2483 2419 4 2384 2193 3 2247 2137 8 2408 2363 2 2249 2172 6 2305 2220 7 2226 2191 6 2 8 3 5

This output contains the

• band_hemisphere identifier ($1234\_2$),
• the shift value ($1111$),
• the fpo reference band frequency ($308.859375\ \mathrm{Hz}$),
• the number of triggers files read ($8$),
• the multiplicity of coincidence found ($N_{coin}=5$),
• arithmetic mean values of the frequency $\bar{f}$ (range $[0:2\pi]$, corresponding to the width of the band), mean frequency derivative $\bar{s}$ (spindown, in $Hz/s$), equatorial coordinate system sky positions $\bar{\delta}$ ($[\pi:-\pi]$) and $\bar{\alpha}$ ($[0:2\pi]$), and the mean signal-to-noise ratio, $\widetilde{\mathrm{snr}}=\sqrt{\sum_i \mathrm{snr}_i^2}$ ($5$ floating-point numbers),
• 8 triplets of the frame number, number of all candidates in the corresponding triggers file, and the number of unique candidates after sorting to unique cells ($8\times 3$ integers),
• frame numbers participating in the coincidence ($5$ integers).

Coincidences larger or equal mincoin (default value 3) are recorded in binary files .coi, separately for each shift, in the -output directory. Each coincidence is a set of following numbers:

where

• $N_{coin}$ is the multiplicity of coincidence (written as one unsigned short int),
• $\bar{f}$, $\bar{s}$, $\bar{\delta}$, $\bar{\alpha}$ and $\widetilde{\mathrm{snr}}$ are the mean parameters of the signal ($5\times$float),
• $\mathrm{fr}_{1},\,\dots\,\mathrm{fr}_{N_{coin}}$ are the frame numbers ($N_{coin}\times$unsigned short int),
• $\mathrm{p}_{1},\,\dots\,\mathrm{p}_{N_{coin}}$ are the positions of candidate signals that took part in the coincidences, in their corresponding trigger files ($N_{coin}\times$int)

in order to recover the information about the original triggers for further studies. .coi files can be read with the auxilary read_coi code:

% gcc -o read_coi read_coi.c -lm 
% ./read_coi 
# num_of_coincidences    mean_val_of_pars (f, s, d, a), snr    frame_num:trigger_num_in_trigger_file
# (see http://mbejger.github.io/polgraw-allsky/coincidences for details)
5 9.956637e-01 -1.108304e-09 -1.125853e-01 1.974630e+00 1.246469e+01 6:777 2:708 8:968 3:701 5:603 
3 9.972902e-01 -1.174760e-10 -1.235594e-01 1.968819e+00 9.154494e+00 8:983 3:693 1:669 

where the pairs frame-number:candidate-position-in-trigger-file have been arranged for readability.