|
|
|
# File formats used by PyFstat
|
|
|
|
|
|
|
|
This page documents the various file formats and how they are used.
|
|
|
|
|
|
|
|
## MCMC Searches:
|
|
|
|
|
|
|
|
### Data output
|
|
|
|
* `outdir/label.log`: On importing `pyfstat`, two logging streams are setup:
|
|
|
|
one stream which is written to the terminal, and a second which is saved to a
|
|
|
|
file `outdir/label.log` where the `outdir` and `label` are given when
|
|
|
|
initialising the search. While the terminal output can be suppressed with the
|
|
|
|
`-q` flag, the file is always written with a log-level set to `INFO`. This file
|
|
|
|
is never overwritten, so can be used to search for changes in the setup or old
|
|
|
|
results.
|
|
|
|
|
|
|
|
* `outdir/label.par`: A parameter file containing the maximum detection stat.
|
|
|
|
value, and estimates of the best-fit parameters. This can be read in with
|
|
|
|
`read_par()` and written with `write_par()`.
|
|
|
|
|
|
|
|
* `outdir/label_saved_data.p`: Upon succesful completion of an MCMC search, the
|
|
|
|
results will be saved to a `python` `pickle` file. This pickle file can
|
|
|
|
subsequently be read back and contains many useful outputs such as the
|
|
|
|
`sampler` object, the `lnprobs` and `lnlikes` from the run, and of course the
|
|
|
|
`chains` themselves. Rerunning a script with different parameters, the pickle
|
|
|
|
is overwritten once the simulation completes, however, a backup is saved with
|
|
|
|
an appended `.old` label.
|
|
|
|
|
|
|
|
### Image output
|
|
|
|
|
|
|
|
* `outdir/label_walkers.png`: The position of all temperature 0 walkers
|
|
|
|
during the burn-in + production stage. In addition, the final panel plots a
|
|
|
|
histogram of the detection statistic from all temperature 0 walkers; if
|
|
|
|
a burn-in period is defined this is computed separately and colored red.
|
|
|
|
|
|
|
|
* `outdir/label_init_i_walkers.png`: The same as the `walkers`, but for the
|
|
|
|
`ith` initialisation stage.
|
|
|
|
|
|
|
|
* `outdir/label_corner.png`: A corner plot of the production samples using
|
|
|
|
the [corner](https://github.com/dfm/corner.py) package. This file is
|
|
|
|
generated by `plot_corner()`.
|
|
|
|
|
|
|
|
* `outdir/label_prior_posterior.png`: A plot showing the prior and a KDE of
|
|
|
|
the posterior, generated by `plot_prior_posterior()`
|
|
|
|
|
|
|
|
## Grid searches
|
|
|
|
|
|
|
|
### Data output
|
|
|
|
* `outdir/label_grid_FS.txt`: Upon succesful completion of a grid search, the
|
|
|
|
grid points are saved in plain text format. The order is set by
|
|
|
|
`get_input_data_array` with an additional column being the output detection
|
|
|
|
statistic.
|
|
|
|
|
|
|
|
### Image output
|
|
|
|
|
|
|
|
* `outdir/label_1D.png`
|
|
|
|
|
|
|
|
* `outdir/label_2D.png`: A 2D contour plot of the detection statitistic over
|
|
|
|
the range of parameters; options exist to flatten higher dimension
|
|
|
|
searches. Generated using `plot_2D()`. |