As of 1.0.0 , we use the catalog (Van Leeuwen 2014 Java tool F2) f2 value here to calculate the error inflation
factor. this is because for some sources, the calculated f2 value is much larger than the
catalog value. E.g., HIP 87275 has a catalog f2 of 65.29, and a newly calculated f2 is using
chi2.sf is infinity. Therefore the error inflation in the catalog is ~7, while the error inflation assuming the new f2 is infinity. We adopt the catalog f2 so as to reproduce the catalog solution and errors.
There are a number of sources that this applies to. So far, I have not found any for which this problem matters -- meaning that as long as I use the catalog f2 value to compute the error inflation factor, then the standard errors on the parameters match that from the catalog. But I have not verified every source.
This issue is by no means urgent, because users can easily verify if this is a problem or not for their source, using htof.validation.utils.refit_hip21_object.
But a suggested solution: one should modify the validation.hip_refit_test script to include the error differences (catalog_errors - computed_errors) for the 5, 7or 9 parameters. Then we can see which sources (if any) for which we cannot reproduce the errors exactly.
As of 1.0.0 , we use the catalog (Van Leeuwen 2014 Java tool F2) f2 value here to calculate the error inflation
factor. this is because for some sources, the calculated f2 value is much larger than the
catalog value. E.g., HIP 87275 has a catalog f2 of 65.29, and a newly calculated f2 is using
chi2.sf is infinity. Therefore the error inflation in the catalog is ~7, while the error inflation assuming the new f2 is infinity. We adopt the catalog f2 so as to reproduce the catalog solution and errors.
There are a number of sources that this applies to. So far, I have not found any for which this problem matters -- meaning that as long as I use the catalog f2 value to compute the error inflation factor, then the standard errors on the parameters match that from the catalog. But I have not verified every source.
This issue is by no means urgent, because users can easily verify if this is a problem or not for their source, using htof.validation.utils.refit_hip21_object.
But a suggested solution: one should modify the validation.hip_refit_test script to include the error differences (catalog_errors - computed_errors) for the 5, 7or 9 parameters. Then we can see which sources (if any) for which we cannot reproduce the errors exactly.