GarryJolleyRogers - Wed Nov 25 2009 - Version 1.9
Parent topic: WebHome

StatisticalMeasureOntology

At the meeting the topic arose why we attempt do define some fundamental UnivariateStatisticalMeasures semantics when SEEK is alsp attempting it... Seriously: how can we combine our efforts? Is a SEEK schema on this topic available somewhere, or could this be added here to the topic? I would much prefer a joint solution!

Or: Does anybody know an external semantic definition sufficiently rich to deal with our problems. Some problems are probably specific, especially that often estimates of range are not statistical estimates but human observer estimates, and that ranges or values may exist of unknown interpretation.

-- Gregor Hagedorn - 21 May 2004

This is not what I'm working on in SEEK but Ferdinando Villa at UVM has the current SEEK ontologies available on his www site http://ecoinformatics.uvm.edu/dmaps/growl/ The Statistical analysis one is the nearest to what you are talking about - but it is expressed in OWL.

-- Main.JessieKennedy - 28 May 2004

I looked at http://ecoinformatics.uvm.edu/dmaps/growl/StatisticalAnalysis.owl (the java applet provides does not work for me...). It seems a useful high level statistical concept ontology, but does not go into any details. We can express that the UnivariateStatisticalMeasures SDD so far has defined belong to StatisticalAnalysis.owl#UnivariateAnalysis and StatisticalAnalysis.owl#DescriptiveStatistics. Does anybody knows about something more specific? Who else might be working on it?

And what shall SDD do? We would be very eager to externalize this knowledge except for two problems a) we use audiences here instead of language - but reducing this language would not kill SDD, and b) it would be really important for us to keep a relationship to the Glossary-style definitions. So this is tied in with the question (a very valid one) whether the entire Glossary concept should be spun off. What do you think?

Another option would be to scrap this, and move it mandatorily into the concept tree. So if multiple sets of statistical measures use a mean, the label for it and the specifications would have to be defined twice. We need to discuss this in a separate topic...

-- Gregor Hagedorn - 28 May 2004

You can also get to the ontologies via the seek cvs site: http://cvs.ecoinformatics.org/cvs/cvsweb.cgi/seek/projects/kr-sms/OWLOntologies/. I think the stuff Fernando has is up to date, but the SEEK CVS repository is the source.

In terms of integrating the efforts, I think that Rich Williams, who is working on these ontologies, is open to input. I'm a fan of spinning the glossary stuff out of SDD - but I'm not sure what impact that would have on its utility. I'm also not sure how you'd refer to the OWL document and element which contains the relevant glossary information. One of these proxy things perhaps?

-- Dave Thau - 3 June 2004