Proc compare with no unique id

[Mikeyjh]

Hi all,

I have a general question abour proc compare.

Is the output from proc compare correct when there is no row level unique id associated with the data but I do use a non-unique column as the ID? For example, if I have an Adverse Event dataset where there is no unique id defined but I use subject number as the ID - is the result of using subject number as the ID dependable? I can also manufacture an ID based on subject number,AE text and start date but this is not guarenteed to be unique 100% of the time.

Is there a way to create an ID across both base/compare datasets to resolve this problem?

[data_null__]

If you are doing QC of ADaM data for example use NO ID variable.  You should be able to create exact duplicate of ADAE and compare row by row.

[Mikeyjh]

Is "NO ID" an option or do you mean dont use an ID at all ?

If no ID is defined how does the compare work?

[data_null__]

Don't use ID statement.  PROC COMPARE will compare record by record as if you said ID _N_; sort of.

If you are doing what I think you are, QCing ADaM data, I would use record level compare.  No ID statement.