Steve: here's the problem. They do 99% of their work at the data browser. They need to be able to enter results for all 740 tests at once, so they want to be able to just tab across a set of columns in the order that matches their instrument printouts. The column headers must be as long as possible so that they make sense and they can keep track of where they are.
I tried to get them to go vertical, with one column for the Tissue name (like "Spleen"), another column for the test name (like Chromium), and another for the value (like 100). Other columns, too, like date test made, patient data, that would go in a header table then. She refused, saying their other tables have 100 or so columns in a spreadsheet-like format, that's what her users are used to, "no way" would they like a vertical format. When I found that RBase's limit was 800 columns, I told that that 747 columns could be handled. ooops. Karen > >>“… 740 tests” > > So, is there any way to explain/argue/enlighten/justify a "more righter" > data model, as I think Jim (Bentley) is implying. > > For example, wouldn't it be better to have a table named, say > AUTOPSY_TEST_RESULTS, defined along the lines of: > > DecedentID (CaseID, or whatever - `PatientID` seems too optimistic ...) > ... > TissueTypeCode -- This allows for multiple tests of same tissue types. > TestTypeCode > TestResultValue > ...
