max error on min stat value - how to?

[Markus Demleitner]

Hi Marco,

On Tue, Nov 05, 2024 at 03:57:16PM +0100, Molinaro, Marco via semantics wrote:
> we need some guidance in identifying
> a UCD term for the following field
>
> "Positive error on the minimum mass"
>
> That would go, e.g., for
>
> stat.error;stat.max
> mixed with
> phys.mass;stat.min

I don't think we should establish stat.error;stat.max as "positive
side of an asymmetric error".  If we want to cover asymmetric errors
(see below), I think we need new atoms.  Tentatively:

stat.error.left | size of the negative side of an asymmetric error bar
stat.error.right| size of the positive side of an asymmetric error bar

I'm sure this can be expressed a lot more graciously.

> stat.error;stat.max;phys.mass;stat.min
>
> that seems quite odd and "unreadable".

Yes... let's not do that, that feels so wrong it just has to be
wrong.

> that drops the upper error limit we'd like
> to represent.

Very frankly, I think I could be convinced that for what UCDs try to
do, a plain stat.error might be good enough for both sides of
asymmetric errors; after all, we don't say whether that's "one sigma"
(if you did a Gaussian fit) or "three sigma" or yet something
completely different either, and that seems even more fundamental to
me.

But then I think at this point I sympathise with your desire to represent
the concept "one-sided-error".

         -- Markus