Parameter file usage
To run Enterprise from a command line using a parameter file on a first pair of .par-.tim files in your data directory, please go to examples/ and run:
$ python run_example_paramfile.py --prfile example_paramfile.dat --num 0
Parameter file options
{x} (a number in curly brackets): a separator, indicating that the following parameters are only for model ‘x’. If we specify more than one model and choose
ptmcmcsampleras a sampler, enterprise is launched in a model comparison mode using the product-space method andclass HyperModelfrom enterprise_extensions.timing_package: a keyword argument of
enterprise.pulsar.Pulsar(), a default option istempo2, another option ispint.paramfile_label: a unique label for the output directory, associated with the given parameter file. The label inside a noise model file(s) is (are) also added to the output directory name.
datadir: a directory with .par and .tim files, or a path to pickled pulsars. In case it is a directory, make sure to have only one .par and .tim file per pulsar, with the same base name.
out: output directory with Enterprise/Bilby results.
array_analysis: whether to run analysis on a pulsar timing array, or on a single pulsar (True for array, False for single pulsars).
noisefiles: a path to .json noise files needed to fix white noise parameters. White noise parameters (EFAC, EQUAD, ECORR) are fixed in case you add lines
efac: -1,equad: -1, andecorr: -1to a parameter file.sampler: choose
ptmcmcsampleror any of the samplers compatible with Bilby. Also, you can add any argument of a sampler as a line in a parameter file (e.g.,AMweightforptmcmcsampler), they are automatically recognized.model_file: a path to
enterprise_warpjson noise model files, one for each model (under {x)}. See examples.psrlist: if provided, only pulsars with names from this text file will be analyzed. A file format is a column of pulsar names.
ssephem: Solar System ephemeris model, the default one is DE436.
clock: a clock argument for enterprise, it is passed to
libstempoorpint(timing packages). At the moment, it is not supported for single-pulsar noise analysis (check the use ofPulsar()inenterprise_warp.py). A default option isNone.fref: reference radio frequency for “chromatic” (e.g., DM) noise, used in
enterprise_models.pyand it can be accessed in your own child class ofStandardModels.
Somewhat less useful parameters:
- overwrite: an option to overwrite an old Enterprise output. It is not maintained at the moment.
- load_toa_filenames: keep a list of raw TOA file names (first column in a .tim file, at least for PPTA data) in a variable self.filenames in a parameter file. It was used for advanced noise modelling.
- mcmc_covm_csv: a MCMC covariance matrix from ptmcmcsampler. The idea is to use a covariance matrix for single-pulsar noise analyses to speed up full-PTA analysis. However, it requires a modification to enterprise_extensions, so it is currently not supported.
- ****
Parameter file also automatically recognizes:
- Priors. Default parameters of prior distributions are set in ModelParams class or its child class where you specify your custom noise models.
- Sampler keyword arguments. I.e., dlogz. They should only be specified after the sampler.