[meas] RFC comment - MD #2,3

[Markus Demleitner]

Hi all,

On Fri, Sep 20, 2019 at 05:24:45PM -0400, CresitelloDittmar, Mark wrote:
> I don't think these are very contentious, so hopefully a single response is
> sufficient.
> 
> "(2) In requirement meas.003: After reading the standard, I think I
> understood what that means, although I'm not sure what the reason for the
> requirement is (let alone which use case it is derived from). Let me try:
> "Each error instance must only be referenced by a single measurement." Is
> that what you mean? If so, why?"
> 
> This means that 'Systematic' error should only show up once for any given
> value.  When you allow errors from multiple sources, it opens the
> possibility that any given source (systematic, statistical, etc) could show
> up multiple times.  We don't want that, I think.  If there are multiple
> contributors to systematic error, we can't distinguish them anyway, so
> might as well consolidate them.

Oh, so source is one of "statistical" or "systematic"?  Ok, yes, I
guess we're helping our consumers a lot with that requirement.
Perhaps saying "A value must not be associated with more
than one each of statistical and systematic errors" makes this a bit
clearer?

But if that's the requirement, I can't understand why Measure:Error
is 1:* any more.  would clients interpret the additional error
objects?  [on my first reading, I had thought these might correspond
to different estimates or so, and so I'm actually happy that I was
wrong].

>> [Markus:]
>> systematic error." So... from my own experience I'd say it would be wise to
>> either say a few words on what's a statError and what's a ranError or, if
>> that's too long, perhaps point to some textbook."
> 
> I wanted to say that these come from the predecessor(s), either STC-1.33 or
> Characterization's Accuracy class, but I'm not finding it.  Anyway, I know
> the 3 have been in the model since at least 2016, where they show up in
> discussions I had with Arnold.
> 
> Anyway.. I have no objection to reducing the set to systematic, and
> statistical only.  This is consistent with what CAOM2 has, though it looks
> like it has 'sys' and 'rnd'.

Great.  I don't think terminology is terribly critical here, so it
doesn't matter much whether we call the statistical error
"statistical" or "random".  However, CharDM (I think) has introduced
"statistical" into several other standards (like obscore, where
there's several *_stat_error columns), so I'd frown a bit if we moved
to "random" now (unless there was a strong reason, that is).

         -- Markus