Open cosimoNigro opened 3 years ago
It's always good to add information, but then it must be clear. Combining systematics errors without knowing their statistical distribution is maybe a risk. Unless we assume (or impose) that they are normal too and somehow indipendent. For example, if you put atmospheric transmission as a systematics without considering that it bias also the energy you can get into trouble.
Hi @micheledoro, I thought the experiments could (optionally) provide a global estimate of the systematic uncertainties on flux points. How the different effects are combined into this estimate I think it is outside the scope of the specs.
I was playing along with SED flux points and radiative models. Having the systematic error specified for the flux points, beside improving the fit, takes properly into account the inter-calibration effects in MWL SEDs.
What do you think of adding systematic errors in flux among the SED error columns? It can be a new column like
{NORM_REP}_sys
or{NORM_REP}_syserr
. The experiment producing flux points and statistical errors might thus add its own estimate of systematic errors. Or one might extend an already existing SED file adding a column following what an experiment claims (e.g. 15% in flux).Of course one could also take into account the systematic errors in energy (i.e. along the x-axis) but I think this is more complex and for the time being the ones in flux are a good start.