Skip to content

Releases: ashleychontos/pySYD

v6.10.0

07 Nov 19:46
Compare
Choose a tag to compare

Release v6.10.0 associated with our accepted JOSS publication

Main changes:

  • Python>=3.8 compatibility across all major platforms (i.e. linux, windows, mac)
  • built-in continuous tests via GitHub Action workflow
  • as well as other stuff that can be updated later
  • need for DOI

v4.1.0

30 Aug 22:59
Compare
Choose a tag to compare

New items:

  • The treatment of white noise: white noise can now be fixed or a free parameter and therefore, the number of models compared for a given target is (nlaws+1)x2 (i.e. 8 models if using 3 or less Harvey models). The white noise can be fixed by using the boolean '-f' (or '--fix') flag
  • The number of Harvey laws: you can now fix or force the number of Harvey laws to use for a given star, in particular when comparing to previous results that might've used a specific number. If '--nlaws' is provided, it will still compare two models between fixed white noise and as an additional free parameter. However, the white noise can be fixed and it will all work together.
  • Getting creative with the software output: the (optional) estimate numax module can be disabled by using the '-x' flag. Similarly, the global parameter estimation of numax and dnu can be disabled by using our '-g' flag (i.e. global) which is True by default. Therefore, it is now possible to run only a background fit for a given target by using '-xg' to turn the other features off. On the other hand, it is also possible to disable the background correction ('-b') and derive parameters from the raw power spectrum.

Changes:

  • The show single hash flag '-s' has changed to '-d' (for display) and '-s' is now used to disable the saving of output data and plots
  • The parameters '--nox' (number of bins to use on the x-axis, 50 by default) and '--noy' (or number of orders to display) for the echelle diagram now actually work! '--noy' is 0 by default and will determine a good number to plot but can be specified otherwise
  • The number of trials for the optional estimate numax module can be changed and plotted for up to 6 different bin sizes. Additionally, one can select with one to use by providing the boolean '-a' flag (i.e. ask the user), which will then plot the output and require a user input

v2.1.1

01 Aug 04:25
0e382c1
Compare
Choose a tag to compare

Release associated with our poster presentation at the TESS Science Conference II. This will also be associated with the arXiv posting (link will be updated on Tuesday, August 3rd).

v1.6.6

01 Jul 00:06
4b29481
Compare
Choose a tag to compare

** NEW RELEASE ALERT ** (version 1.6.6)

The main changes associated with the release are enumerated below:

  • Single flags (CLI options with a single dash) are now reserved for a special functionality that is based on Unix operating systems. They can be combined together to shorten the number of arguments used. An example is shown below. Because this functionality cannot have inputs with the associated flags, these are most limited to boolean type inputs (-s for show, -v for verbose, -k for Kepler artefact correction, -x for turning off the first module and -b for turning off the second module):
    • pysyd run --star 1435467 -show -verbose can be shortened to pysyd run --star 1435467 -sv but will do the same exact thing.
  • The simplest model (aka flat white-noise term only) has been implemented BUT this has not been tested on null detections yet. I am curious to see if this works as nicely as I hoped it would.
  • The metric used for model selection is now the BIC, with an option to use the AIC as well. The scipy.chisquare function has been removed, which by default should eliminate the issue with the scipy versions.
  • The background-corrected PS is now saved to a text file in the same directory as the other output data
  • There is a new Constants class available in utils.py that contains all of the constants used in any calculations throughout the software (and therefore easily importable)
  • The plotting “block” has been turned off, so you no longer have to close out of the figure to continue running the software
  • The white noise estimation module has been fixed and does not require the nyquist to be defined. I think this * should * fix the x0 infeasible error we were seeing, which was because the estimated white noise was a nan.
  • CLI options have been further separated into relevant sections so they are easier to know where they are applicable (i.e. pysyd run --help will output text that is much easier to follow)
  • Along that line, many new command line arguments/options have been added, including (but not limited to):
    • the width (--ew, --exwidth) of the power excess region (this is a fractional value of the width calculated based on numax and the numax/width of the Sun). So by default, this is 1.0 but can be any fractional value (i.e. 0.5, 2.0), which will still remain centered on the estimated numax.
    • you can be even more specific and provide the exact bounds to use of this region for the power excess by specifying the upper (--up, --upperp) and lower (--lp, --lowerp) bound (i.e. not necessarily centered on numax). this is helpful if you have power excess near a strange artefact (or anything), where selecting strict bounds helps reduce the noise in the computed ACF.
    • --thresh, --threshold is the fractional value of the FWHM used in the ACF to find the best dnu. Similar to the excess width option above, it’s 1.0 by default but can be adjusted based on the ACF. An example to justify this would be if the calculated FWHM encapsulates two peaks and it’s latching on to the incorrect peak.
  • Some of the modules have been modified but the functionality is still the same. I’ve attempted to break down some of the functions into more intuitive steps, the end goal being a more modular software package.
  • The harvey models are now directly sampling in {tau, sigma} space, as opposed to the original {a, b} parametrization. This helped ensure the priors were strictly physical values.
  • Due to the numerous upper/lower bounds used throughout the software, the names/keys for these are slightly more descriptive. For example, lowerx/upperx is now lower_ex/upper_ex and same for the background fitting routine, so it’s clear that these are general bounds for the modules. Lower_ps/upper_ps is for defining the region with the power excess (optional), lower_ech/upper_ech is for defining the region in a folded PS (i.e. echelle diagram) for whitening mixed modes from the PS before attempting to estimate dnu (optional). NOTE: this may require you to either re-run pysyd setup or to change the keys (i.e. columns) in the info/star_info.csv.