Skip to content

DRAFT: new file formats

Karl Gordon edited this page Apr 1, 2019 · 3 revisions

Overall Thoughts

More robust file formats needed.

More information - all the details of the run so that what was done can be easily recovered.

Fold in the needs of the MegaBEAST.

Add unit info even if we don't use it yet.

Log file

Need to log the input information including the priors

Observed Catalog

Specify the names of the columns and units. How are the filter names specified? Probably should adopt the long form throughout the BEAST (e.g., 'HST_WFC3_F475W' instead of the short form of 'F475W' given the ambiguity between WFC3 and ACS, for example).

Artificial Star Test Inputs/Outputs

Specify the names of the column and units.

Evolutionary Tracks/Isochrones

Astropy table compatible format that includes unit information. FITS/ASDF.

Stellar Atmospheres

Specify the format of the stellar atmosphere grid files. Units would be useful.

Physics Grid

Ideally, something that can be read with astropy.tables or hd5. Speed is an issue given the size. FITS/ASDF/HDF5. Compression?

Has to save the filter names in an easy to find format (make it easy to remove filters from fitting).

Priors saved as columns separated between grid weights and prior weights (and a combined version).

Observation Grid

Ideally, something that can be read with astropy.tables or hd5. Speed is an issue given the size. FITS/ASDF/HDF5. Compression?

Toothpick: Bias and noise for each model.

Trunchen: Bias and covariance matrix for each model (not full nxn matrix as symmetric, decompose?).

Results

Would be good to combine the current 3 files into a single file to keep the information together.

Includes summary statistics, 1D posterior PDFs, and (sparse) nD likelihoods.

Heterogeneous data means HDF5/ASDF.