Bilby Analyses¶

This section of the cookbook provides several examples which you can adapt for setting up analyses using the bilby pipeline.

Because asimov is configured to always expect that noise estimation is handled by a different pipeline, bayeswave, it’s probably sensible to have a look at its section in the cookbook first. For the analyses on this page we’ve always called the PSD job psd-generation to be consistent with the recipe in the Bayeswave page, but if you’ve called it something else you’ll need to update it in the dependencies.

Getting started¶

In order to set-up a Bilby job you’ll need to set up some defaults which will apply to every Bilby analysis in your asimov project, unless you specifically set things differently in one of the analyses. These defaults set up things like the configuration for the scheduling system on the compute cluster.

You can see example defaults for all pipelines used by LIGO analyses here, but below are the set for Bilby on their own:

kind: configuration
pipelines:
  bilby:
    quality:
      state vector:
        L1: L1:DCS-CALIB_STATE_VECTOR_C01
        H1: H1:DCS-CALIB_STATE_VECTOR_C01
        V1: V1:DQ_ANALYSIS_STATE_VECTOR
    sampler:
      sampler: dynesty
    scheduler:
      accounting group: ligo.dev.o4.cbc.pe.bilby
      request cpus: 4

If you save this file as bilby.yaml you can then apply these defaults to your own project by running

asimov apply -f bilby.yaml

You should only need to do this once, and if you’ve already set them through e.g. one of the files containing all the LIGO settings you don’t need to worry about this stage.

Standard binary black hole analysis¶

The following yaml file will allow you to analyse an event using the normal setup required to analyse a binary black hole signal, and using the IMRPhenomXPHM waveform model.

kind: analysis
name: Prod1
pipeline: bilby
waveform:
  approximant: IMRPhenomXPHM
comment: Bilby BBH parameter estimation job
needs:
- psd-generation

Standard binary neutron star analysis¶

The following yaml file will allow you to analyse an event using the most straight-forward setup for analysing a binary neutron star signal. Note however that this is not the fastest way to perform this sort of analysis, and more efficient methods are covered below under ROQ methods.

kind: analysis
pipeline: bilby
name: bilby-bns
needs:
  - psd-generation
approximant:  IMRPhenomPv2_NRTidalv2
comment:  IMRPhenomPv2_NRTidalv2 analysis
likelihood:
  frequency domain source model: lal_binary_neutron_star
sampler:
  sampler: dynesty

You can see that compared to a BBH analysis we need to specify some additional information about the type of likelihood function and waveform generator we want to use.

ROQ methods¶

We can achieve considerable speed improvements by employing reduced order quadrature bases in our analysis, especially when using binary neutron star waveforms. Some additional care must be taken with these, especially when choosing the prior range, as these are not valid throughout parameter space.

The YAML file below sets up a 256-second long ROQ analysis, which would be suitable for a BNS signal such as GW170817. However, you should double-check that the prior this sets isn’t wider than what you desire; if it is you should adjust it accordingly, provided your desired prior is within the boundaries defined below.

Because the datafiles for these runs are stored on the LIGO computing clusters you’ll need to update the filepaths to the locations where you’ve downloaded a copy of the bases to.

kind: analysis
pipeline: bilby
name: bilby-roq
needs:
  - Bayeswave
approximant:  IMRPhenomPv2_NRTidalv2
comment: IMRPhenomPv2_NRTidalv2 256s ROQ job
likelihood:
  marginalization:
    phase: True
  frequency domain source model: lal_binary_neutron_star_roq
  calibration:
    sample: True
  type: ROQGravitationalWaveTransient
  roq:
    folder:  None
    linear matrix: /home/roq/IMRPhenomPv2_NRTidalv2/bns/basis_256s.hdf5
    quadratic matrix: /home/roq/IMRPhenomPv2_NRTidalv2/bns/basis_256s.hdf5
    scale: 1.0
sampler:
  sampler: dynesty
priors:
  default: BNSPriorDict
  chirp mass:
    minimum: 0.92
    maximum: 1.70
  spin 1:
    maximum: 0.4
  spin 2:
    maximum: 0.4