...where in Table 7 of \cite{Madsen_2015} we allowed both the Slope and Offset to vary as a function of time. Values were stored for the 2012 and 2015 epochs and the NuSTARDAS pipeline then interpolated the corrections to a given epoch for an observation.
In subsequent analyses of observations of the Crab we noted allowing the Offset parameter to vary with time was artificially introducing low energy residuals in the Crab spectrum. We instead re-fit the 2012 and 2015 calibration source data with the Offset value frozen to zero, thereby forcing all time-dependent variations into the Slope parameter. This remove the artificial low-energy residuals in the Crab spectrum. The updated values for the 2015 epoch calibration source data are given below. The method we use to determine the offsets is formally limited by the digitization of the NuSTAR energy channels (which have a width of 40 eV). To achieve a 20 eV shift (enough to shift a count from one channel to the next) at the 86.54 keV line energy of the calibration source, the Slope must differ from unity by at least 2e-4 so all values in the table below are assume to have an uncertainty of (2) in the last reported digit.
The implementation of the gain CALDB files assumes that the variations are linear with time with the measure changes in the gain corresponding to a drop of 0.2% per year from 2012-2015. This does not currently extrapolate beyond 2015.