I tried the sort with nodupkey and it didn't get me what I was looking for. Hold off the above question, I didn't phrase it correctly. Paul ... View more

Hi Patrick I tried part of your code for alpha characters above, which is definitely beyond my SAS capabilities--in spite of the fact I have been programming with it for 3 years, and it worked great for the docket numbers! I pasted the modified version of it below. My question is can this be modified to dummy a first, last, middle name field as in the below format? Smith, John Doe, Jane, V. Jackson, Randy, G. Hanson, Therese data masked_docket_lookup;   set af3(keep=docket);   _masked_docket=docket;   stop; run; %let rc=byte(65+floor(ranuni(0)*26)); %let rd=floor(ranuni(0)*10); data want(drop=_:);   if _n_=1 then     do;       if 0 then set af3 masked_docket_lookup;       dcl hash Hdocket(dataset:'masked_docket_lookup');       _rc=Hdocket.defineKey('docket');       _rc=Hdocket.defineData(all:'y');       _rc=Hdocket.defineDone();     end;   set af3 end=last;   if Hdocket.find() ne 0 then     do;       _masked_docket=cats(&rc,&rc,'-',&rd,&rd,&rd,&rd,&rd,'-',&rd,&rd);       _rc=Hdocket.add();     end;   docket=_masked_docket;   if last then Hdocket.output(dataset:'masked_docket_lookup_tmp'); run; ... View more

Hi Jwillis I think something like that might work for our docket numbers because there are a limited number of prefixes. Per Reeza above, someone could decode this, but it is going to a specific vendor who is signing a confidentiality agreement (and not available to the public), so I don't think this is a risk. Let me see if I can come up with something like the above and test it. Paul ... View more

It can be made up, but it has to look real. For example, our dockets are in the format of NN-20912-13. This can't be dummied to say something like 4830989453XXXXX. Dummied, it would be to look something like JH-84092-02. And I can't make up hundreds of thousands of records for these values.. It has to be some kind of systematic dummy version of the original. The variables themselves are things like docket numbers (see above), first/last names, sequential database IDs. Paul ... View more

The problem is there are thousands of unique values for each of the variables. Unfortunately, its not like a gender variable where there is a manageable number. Paul ... View more