[QUANTITY] Do we need it? Some quantities
martin hill
mchill at dial.pipex.com
Mon Oct 27 14:40:36 PST 2003
Being a bit radical here, but what is quantity for? I've read the list (honest)
but it strikes me that really a quantity is simply a number. A range for
example presumably consists of two quantities. So perhaps we should start
further back and look at what kinds of quantities we want to describe -
personally I suspect we'll end up with a bunch of different 'quantity models'
some of which are not necessarily related...
So as a non-astronomer, we might have:
Value: simple number (integer or real)
Error:
simple (+/- absolute)
percentage (+/- percentage)
normalised* (sigma?)
Range: Min Value + Max Value
might have errors attached to range or to values?
Value & Range might have a unit attached, but only a simple error will.
*not sure of the right term here
Including quality is a scary topic (for me as an implementer!). Can we make it a
separate [Quality] model altogether? It will get attached to all kinds of
things, including presumably complete data sets?
Martin
Quoting Brian Thomas <brian.thomas at gsfc.nasa.gov>:
>
> Hi all,
>
> I came away from the ADASS/IVOA meeting being encouraged that some
> progress could be made at data modeling as long as we tried to side
> step
> semantic issues of naming and coding specifics, and instead started
> at
> the most general level of needs (or requirements). This means no
> attempts
> to produce "the" UML diagram with all specifics and trying to hammer
> each
> of our models home. The details in each are too much for the others
> to
> generally comprehend, or even, in general, want to spend time
> studying.
>
> Hence, I propose that we attempt to gather requirements for the
> quantity, to
> better define what it is we all want. These requirements would be
> *then* used
> to create a consensus UML diagram *later*. I would hope that for this
> discussion,
> we can just debate the requirements of "quantity", perhaps with
> points illustrated
> via a UML diagram or a use-case, but no "overall" UML diagram being
> produced
> until we all largely feel the "fundamental" requirements have been
> generally
> agreed to.
>
> If this experiment is successful, I would hope that the DM group
> will
> recommend a draft of these requirements as part of the greater
> "Observation"
> whitepaper that I understand Jonathan wants to prepare, or perhaps as
> a
> separate IVOA note.
>
> So, to start the ball rolling, here's some requirements that I saw
> in the various models (all requirements start with the "="), and/or
> popped
> up the the informal discussion at IVOA/ADASS. I have attempted
> to arrange these in order of general acceptance, rather than
> "importance":
>
> <requirements>
>
>
> = Quantity is a container class that holds scientific, engineering
> information.
>
> = Components of the observation model will inherit, as needed, from the
> quantity.
>
> = The quantity has an associated class, "coordinate transform/mapping",
> which is
> used to transform one quantity into another (dimensionality of the two
> quantities
> is the same).
>
> = The information in quantities, in order to make it valuable and machine
> readable
> requires that it be described by meta-data that include type of units,
> type of data
> format (such as "long", "float", "string", etc) and its accuracy (which
> includes
> things like "quality" flags and "statistical/systematic errors").
>
> = The quantity may be multi-dimensional (this is _almost_ a universal
> feature of extant models).
>
> = The quantity may hold information which comprises scalars, vectors or
> other quantities (tuples).
>
> = Primary/top requirement: The quantity will be used to facilitate the
> search,
> exchange and data fusion in the VO. [The "data fusion" part leads to the
> 3rd requirement
> in the list]
>
> = The quantity will be able to completely describe all the information
> (data or
> meta-data) in a FITS file or VOTable.
>
> which (I think) leads to..
>
> = The quantity is a container class for _meta-data_ as well as _data_.
>
> </requirements>
>
>
> So there is a "starting" list. Nothing "official", but perhaps it can
> lead to
> some consensus document that is.
>
> Thats all for now,
>
> Regards,
>
> =b.t.
>
> --
>
> * Dr. Brian Thomas
>
> * Code 630.1
> * Goddard Space Flight Center NASA
>
> * fax: (301) 286-1775
> * phone: (301) 286-6128
>
>
>
>
--
Martin Hill
07901 55 24 66
www.mchill.net
More information about the dm
mailing list