[QUANTITY] Are we agreeing?!?

[Gerard Lemson]

Hi
Let me give you some of the reasons why we separated out Quantity from
Measurement, which
will at the same time explain why we use these names.
The main one is that we wanted to make the concept of Measurement explicit,
that is, we wanted
within our model, to be able to describe that a measurement, in the usual
meaning of that word,
had been performed. A measurement is where you try to assign a value to some
property (a Phenomenon)
following a well described "Protocol". All that information, *how* the
measurement was performed, is
very important for being able to interpret the outcome of this measurement.
We believe that outcome
to be a Quantity, that is a [number,unit] pair or some generalization of
that. Furthermore and separate from that,
we associate a measurementError (another Quantity, in general of a different
type) to the Measurement itself,
For it is because of the protocol that was followed in the measurement that
a certain error is introduced.
Note btw that this is not the only place where an error can be assigned.
There are many cases where the
interpretation of one measurement is dependent on a series of other, related
measurments. For example the errors
in individual "pixels" of CMB maps are in general highly correlated. To take
this into account we group individual
measurements in what for lack of a better name we called the "Result", as it
is the result of an Experiment, which in general
contains a number of Measurements. It is this whole result which in the end
allows one to interpret of all the individual
quantities.

What this implies to us is that Measurement is not a subclass of (our
concept(ion) of) Quantity. The relation between
the two is not an "is-a" but a "uses" relation. We believe that one should
be extremely careful using the "is-a" relation.
Calling everything a Quantity removes any meaning from that concept, turning
it into something like Java's java.lang.Object.

One more very important observation about our model (see
http://www.g-vo.org/materials/UDM-Poster.pdf
for a version of the poster, which may not show nicely on all browsers
though).
We do *not* want to imply that the part in our model currenly dealing with
Quantity is the only way in which quantities should
be stored or sent around in VOTables or so. Our goal was to try to describe
what the context is of a Quantity. What more
information is required to make sense of an individual number plus unit plus
... .
We think that a model is required that describes the storage of quantities
in archives. This imo is the domain of
Brian and Ed's model. We try to supply thoughts on what other information,
currently in general not explicitly stored in
archives, has to be supplied by datacenters to publish their data in a
uniform manner.

Our Quantity as it appears in our current model is really an instance of a
single Phenomenon.
Phenomenon is what SI called a "quantity in the generalized sense" (see
http://physics.nist.gov/cuu/Units/introduction.html).
Our Quantity finally is what there is called a "value of a physical
quantity".
We wanted it to correpsond to the smallest unit of measurement one may
conceive.
So individual pixels in an image are Quantities, the whole image is a
result.

Enough for now, if not too much.

Bye
Gerard



--
* Gerard Lemson                       * Tel: +49 (0)89 30000-3316
*
* MPI fuer extraterrestische Physik   * Fax: +49 (0)89 30000-3569
*
* Giessenbachstrasse                  *
*
* Postfach 1312                       *
*
* D-85741 Garching, GERMANY           * email: gerard.lemson at mpe.mpg.de
*


> -----Original Message-----
> From: DIDELON Pierre [mailto:dide at discovery.saclay.cea.fr]
> Sent: Thursday, October 30, 2003 2:48 PM
> To: patrick.dowler at nrc-cnrc.gc.ca; thomas at mail630.gsfc.nasa.gov;
> brian.thomas at gsfc.nasa.gov
> Cc: Gerard.Lemson at mpe.mpg.de; shaya at mail630.gsfc.nasa.gov;
> jcm at head-cfa.cfa.harvard.edu
> Subject: Re: [QUANTITY] Are we agreeing?!?
>
>
>
> > From brian.thomas at gsfc.nasa.gov Thu Oct 30 14:35:53 2003
> > From: Brian Thomas <brian.thomas at gsfc.nasa.gov>
> > To: DIDELON Pierre <dide at discovery.saclay.cea.fr>,
> >         patrick.dowler at nrc-cnrc.gc.ca, thomas at mail630.gsfc.nasa.gov
> > Subject: Re: [QUANTITY] Are we agreeing?!?
> > Date: Thu, 30 Oct 2003 08:35:46 -0500
> > User-Agent: KMail/1.5
> > Cc: Gerard.Lemson at mpe.mpg.de, shaya at mail630.gsfc.nasa.gov,
> >         jcm at head-cfa.cfa.harvard.edu
> > MIME-Version: 1.0
> > Content-Transfer-Encoding: 7bit
> > Content-Disposition: inline
> >
> > On Thursday 30 October 2003 03:59 am, DIDELON Pierre wrote:
> > > From Pat's diag. Error is "included" in measurement.
> > > So is there add. info included in Quantity? Or are you
> refeering to "your
> > > quantity" which corresponds to "pat's measurement"?
> >
> > 	Yes, thats right. Its largely semantics as to whether "measurement"
> > 	is a "quantity".
> >
> > 	To help resolve this problem, as I wrote to Pat, I suggest
> that we consider
> > 	that we *always* need to make some statement about the
> errors. This is
> > 	to say, we need to be clear when they are present and when
> they are not
> > 	present. When errors are _not_ present, I see at least 2
> cases that we need
> > 	to clearly describe: 1."no errors because this number is
> exact" (e.g. a "constant" or "defined"
> > 	value) AND 2."no errors because its not defined..Im lazy"
> (this happens *alot*
> > 	in the data that I have seen). Both "no error" cases give
> separate meaning to
> > 	the "value" and are important.
> >
> > 	Therefore, even if you dont like re-naming "measurement" to
> "quantity"
> > 	(which I prefer), it should be clear that we *have* to
> group the values with
> > 	errors at the "base" DM level. This means that at the
> least, "measurement"
> > 	class is part of the base package.
> >
> As long as the agreement is understand by all parties, I have no
> objection.
> But how did you call then pat's quantity class?
> It has basic usefull functionnalities and can be used on its own, I feel.
> Just to be sure that we spoke the same langage (even if for me it is not
> completly true) and use the same words for the same abstraction/concept.
> Sincerely,
> Pierre
>