Re: One-to-many database with merge requirement

You likely should includes some kind of alias table. Here you'd have the identifier table, the results table and an additional alias table that allows mapping one identifier to another. You'll still likely require human intervention to build the aliases, but you need some way of relating them once you do the work. The only reason for keeping the identifier separate is the chance that someone does use the same identifier. BTW, you might want to ask for birth date and/or phone number. Having those should dramatically improve your chances of matching identities...

> I am leaning toward a model with one record per test, with identifying
> information included in each record. It would then be up to a human
> operator to decide which records actually refer to the same person when
> doing test-retest comparisons.
>
Received on Tue Apr 18 2000 - 00:00:00 CEST