Kevin wrote:
It strikes me that
- in some ways these two suggestions are very similar (whether the
thing goes into the taxon x character cell or is attached to it is immaterial from a data point of view). 2. the real difference is in where their values are defined. Gregor's "special states" are defined in the instance document, while Steve's flags are (I assume) defined in the schema. Does this make a difference?
Two differences I see:
a) The Brazil proposal allows the project designer to constrain which "missing data indicators = special states = CodingStatus" are available within a given character while scoring this character in an item description.
This is why the "special states" occur indeed within a character definition: They are not defined there, but by referring to the global/shared definition level, they are _enabled_.
What I find nice is that this uses the same kind of mechanism the inheritance of shared state sets uses: either inherit the full set for a given character, or restrict it to certain states.
BUT: it is not necessary to do it like this. This is truly just my private attempt to get some structural generalization into the model.
Also: I tried to raise the question of how important it is to be able to enable/disable certain special states. I find it useful if the list of special states has, say 5 options and you want to produce a web interface, but I am uncertain whether it is worth it. If it is a side-effect of having a structurally simple model, I welcome it, but otherwise I can just as well live with "missing data indicators = special states = CodingStatus" always enabled for all characters if we can reach a consensus about that.
b) Second difference: special states do need a label/wording whenever they are reported. To achieve multilinguality this wording must be user-definable (i.e. by the designer of the project/the terminology). To me this points into the direction of putting it into the xml instance document, handling it like the state definitions. Again other solutions are possible.
I am not fully happy with the way the Brazil model works, and it obviously confuses people.
Perhaps one of the confusing things is that
1. SDD makes definitions 2. Project designer makes global/shared definitions 3. Project designer creates characters, referring to shared definitions (point 2 above), inheriting the properties of 0-several shared definitions, PLUS in the case of categorical states defines local states (but in the Brazil model not in the case of "statical parameters = global measures" or "special states") 4. The scientist using a designed project scores the descriptions based on the decisions of the designer of the terminology.
Because SDD is able to constrain things through the schema, this 4 layer model is acutally NOT represented in the Brazil model (and the model could perhaps be made more understandable by changing that? I really don't know!). Instead, SDD definitions like "Special states" (or whatever better to call them) and "statistical parameters" are treated like point 2, but the designer has only limited choice. The xml schema requires the designer to have a given number of shared definitions for special states and statistical parameters present in that part.
However, this contraction allows the designer to add labels to these required definition in new languages. I currently do not see a better solution for this, but I welcome any suggestion!
Can anybody help me in making the explanations above or the questions poised in "stepping back" more intelligable?
Gregor ---------------------------------------------------------- Gregor Hagedorn (G.Hagedorn@bba.de) Institute for Plant Virology, Microbiology, and Biosafety Federal Research Center for Agriculture and Forestry (BBA) Koenigin-Luise-Str. 19 Tel: +49-30-8304-2220 14195 Berlin, Germany Fax: +49-30-8304-2203
Often wrong but never in doubt!