SED and Binned Likelihood Formats

[woodmd]

This is an open thread to collect comments and feedback on the proposed formats for Binned Likelihood Profiles. Since these formats are effectively just SEDs with extra information in each bin it would probably be a good idea to try to share conventions with the Flux points format. @zblz Did you have any thoughts on this?

Here are a few open questions:

• Do we want a fixed 95% CL for the UL columns? I saw that the naima format specifies the confidence level with a header keyword so maybe we should do something similar.
• Do we need a column to specify the bin center? The current format only specifies the lower and upper edges (E_MIN and E_MAX). Differential quantities are assumed to be evaluated at the geometric center of the bin but we could add a bin center column (E_CTR) that would allow the evaluation point to be explicitly defined.
• Which columns do we want to make optional? The current specification calls for five different unit conversion factors (DFDE,E2DFDE,FLUX,EFLUX,NPRED). I mainly did this for convenience so that one doesn't need to know anything about the underlying spectral model to convert between units. Of these E2DFDE could clearly be optional (or maybe dropped entirely) since it can be trivially computed from DFDE. NPRED provides additional information that can't be rederived from the other four unit columns but since it may not always be available it should probably also be optional.
• Do we want a column for exposure? This can be inferred from FLUX and NPRED but having it as an independent column could be useful in some circumstances.
• Do we need a column that describes the model for the spectral distribution within each bin? For instance when making LAT SEDs we typically substitute the source with an index=2.0 power-law. For many applications this doesn't really matter (e.g. when using narrow bins) but when using wide bins or when the effect of energy dispersion is significant the assumed shape within the bin can have a big influence on the measured flux. One option would be a column for the logarithmic slope of the bin model evaluated at the bin center.