Re: [Gnumed-devel] Normcurves, again

[Christof Meigen]

Hi Karsten and others,

[Richard Terry:]
> > pardon my ignorance but how to I get these 
> > two files to 'run' or do something 
> > so I can comment?

The .DTD is commented and should be human-readable It's a "document
type definition" which defines a certain type of XML-documents, here:
XML documents which contain medical normcurves.  the file testnorm.xml
is a sample for how such a file looks like.

>From the DTD it is pretty obvious how a normcurve must look
like. There are certain things I left out, like

-> "Ethnic" parameters as sometimes provided in the US 
   I couln't think of a resonable and pc way to do it, 
   and "race-mixing" makes this questionable anyway.
   Instead a Country Code is provided.
-> Normcurves for patients with certain diseases,
   for example Turner-girls. We don't need that here
   in Leipzig, but maybe you do? Raise your hand.
-> Super-fancy stuff I've nerver heard of, 
   like strange distributions of values (other than
   (skew)normal), three-dimensional growth charts
   ....

Other things are in there, for example the possibility to enter
normcurves with an X-Axis other than age (like in "weight for
height"-charts for babies) or the possibility to use smooth splines
instead of some fixed values for some ages.  Maybe no one needs that?

Once there is a - more or less - consensus about what should be in it
and what not, I'll start to port my Java and C-Routines to Python to
provide a way to actually _use_ the norms within GnuMed.


> Christof is thinking that in some offices there may not be a
> server or if there is there may not be a Linux server
> worthy capable of safely running PostgreSQL. Therefor
> he wants to store reference data in XML files.

Uh, no. While indeed this opinon came up during our discussion, it was
raised by Torsten Menzel, not me. I wouln't want to change the way
GnuMed saves its patient data etc.

Norm curves, however, seem to me more like some kind of "configuration
file". I chose XML because

- It's standard. It's well structured. 
- It's text. You can send it via Email or download it from
  the web and you immedialtly have it ready to use.
  As mentioned, I imagine that gnumed.org may provide
  a normcurve-search-engine (for example, when you as a physician 
  have some new vietnamese patients, cou can look up
  normcurves for them). With XML, you can simply output
  the XML-stuff and say "copy and paste this into your
  normcurve-file". When the normcurves are stored
  in a database, things become more complicated.
- In Python, there is a built-in parser for XML
- It can be used outside of GnuMed, for example by researchers,
  who still lack a standard format for normcurve-interchange

Furthermore normcurves don't fit well into an relational databese
scheme, while they can be naturally expressed in XML.

Still, if the most of the GnuMed people think a SQL-based storage
scheme would fit better into the "philosophy" of GnuMed, I'd also do
it that way.

        Christof