Defined field for uncertainties on each axis #53
Comments
|
@Chantler1 Thanks -- I agree that adding fields to hold uncertainties for all the signals (including energy) is a good idea. I believe that this can be solved at the level of documentation, convention and "recommended" settings (that is, that the code already can support this). The XAS data library for the database uses names like "energy_stderr", "itrans_stderr", etc. We should probably use the same naming conventions. I would propose we add those as recommended labels for arrays containing standard error (or, at least some value expressing the estimated uncertainty in the corresponding data), but another naming convention could be fine too. I think we haven't explicitly mentioned units for any of the intensity values, or distinguished attenuation from absorption in the intensity signals. I think it's OK to allow for these, but almost all the data will not make this distinction. Again in the data library code, there are "energy_notes", "i0_notes", etc. These could be used to hold such information, but probably this would be only "human readable". Thinking bout ow to encode these better would be interesting.... |
|
Thanks for that Matt Chris Christopher Chantler, Professor, FAIP From: Matt Newville [notifications@github.com] @Chantler1https://github.com/Chantler1 Thanks -- I agree that adding fields to hold uncertainties for all the signals (including energy) is a good idea. I believe that this can be solved at the level of documentation, convention and "recommended" settings (that is, that the code already can support this). The XAS data library for the database uses names like "energy_stderr", "itrans_stderr", etc. We should probably use the same naming conventions. I would propose we add those as recommended labels for arrays containing standard error (or, at least some value expressing the estimated uncertainty in the corresponding data), but another naming convention could be fine too. I think we haven't explicitly mentioned units for any of the intensity values, or distinguished attenuation from absorption in the intensity signals. I think it's OK to allow for these, but almost all the data will not make this distinction. Again in the data library code, there are "energy_notes", "i0_notes", etc. These could be used to hold such information, but probably this would be only "human readable". Thinking bout ow to encode these better would be interesting.... — |
Dear Matt and Bruce
I feel that we should have definitions for column fields for uncertainty (standard deviation or standard error) both for the uncertainty in mass attenuation coefficient (cm^2/g) and also for the energy.
While energy is sometimes represented as a single offset with a single energy uncertainty, ideally it maps into a field for each channel, whether in slew or step mode.
For the attenuation coefficient it is quite clear that independent uncertainties permit a more unbiased determination of quality, especially with metadata indicating methodology or provenance. It also permits strong hypothesis testing. Hope that this helps ;)
In principle I would also expect a defined column field for both attentuation and for absorption. The latter is of course more directly relevant for XAFS, but most users will not know how to extract it so the attenuation column represents more fairly the data.
Very best wishes
Chris
The text was updated successfully, but these errors were encountered: