[NeXus-committee] RFC variants and esd

[Pete Jemian]

Wrote this comment too quickly.

On 10/24/2013 12:50 PM, Pete Jemian wrote:
> Fundamentally, a field would have an attribute that names a child group
> in which the uncertainty (single or plural) is described with fields.

Look at the figure below this URL:

A field with numerical data (such as "I") has an uncertainty attribute 
that gives the name of another field (such as "Idev") in the same group 
containing numerical value(s) for the uncertainty.  The name of the 
uncertainty field is specified by the user.  That field contains the 
estimated uncertainty associated with the first field (such as I +/- 
Idev).  For fields with multiple uncertainties or other complexities 
that contribute to the uncertainty, a subgroup, perhaps an NXnote, can 
be created to contain those fields.  The name of that subgroup is given 
in the components attribute of the uncertainty field.

In the example shown, I is the measured intensity, Idev is the estimated 
uncertainty and has several contributions each of which are documented 
in the I_uncertainties group: Johnson noise, shot noise, and estimated 
standard deviation (esd).  For the purposes of data analysis, I and Idev 
are used and the rest are likely ignored.

The example following this suggests a way to represent multiple 
uncertainties such as standard deviation and standard error.  It does 
not recommend a way to handle that data, just how to represent it. 
Suggested is that the first item in the list is the one to use in analysis.

Pete