On 03/12/2013 15.10, Noble, Richard wrote:
To what extent does the supposed significance of these statistics depend on an assumption of consistency and non-duplication in the database? Has this really more to do with features of records than with the entities they represent?
That’s an honest question. My own work exposes me to wide disparities among levels of description generally, often exhibited among as many as two dozen duplicate records for the same entity; but that could well be of minimal statistical significance in the database as a whole.
Excellent points. The reason I played with these numbers is to get an idea of how many versions there are, and if catalogers are supposed to make the RDA relationships useful for the public (so that people really and truly can find “adaptation of”, “parody of” and so on), to have at least some kind of idea how many records will have to be touched if we are to get any thing resembling “consistency” in the database. Otherwise, any searches can never even hope to succeed and it seems hard to justify the time and effort going into adding these relationships.
As you point out, there are lots of problems with this analysis: there are tons of duplicate records in Worldcat; I mentioned that catalogers may not add information, e.g. 240 when they should; including the 130 field into the mix is highly problematic. But otherwise, we seem to be left with no idea whatsoever for what it would take to make the FRBR/RDA relationship fields useful. Perhaps others will have ideas on how to get some really useful numbers.
The RDA people have never addressed this issue so I thought I would give it a start.
Then we can think about the subfield e.(!!)