It seems that there are currently two parallel error estimation pipelines in hera_pspec. The first is inside of pspecdata where a user provides a model of the data covariances and is provided propagated power spectrum error bars and covariances. The second is in utils and includes a completely separate system for computing thermal noise from autocorrelations and/or calculating error bars from bootstrap sampling over redundant groups of baselines. Both methods in utils are currently run as a postprocessing steps and are kind of easy to miss when there is parallel error bar functionality in pspec_run.
I think it would be clearer if we integrated these functions in utils into pspec_run, adding options to perform them as part of the primary power-spectrum calculation.
It seems that there are currently two parallel error estimation pipelines in
hera_pspec. The first is inside ofpspecdatawhere a user provides a model of the data covariances and is provided propagated power spectrum error bars and covariances. The second is inutilsand includes a completely separate system for computing thermal noise from autocorrelations and/or calculating error bars from bootstrap sampling over redundant groups of baselines. Both methods inutilsare currently run as a postprocessing steps and are kind of easy to miss when there is parallel error bar functionality inpspec_run.I think it would be clearer if we integrated these functions in
utilsintopspec_run, adding options to perform them as part of the primary power-spectrum calculation.