In addition, let's move away from the actual code for a second and explain the analysis you are trying to do from a higher level, with no discussion of code. You have data, you want a certain result, what is that desired result (without mentioning code)? Show us the desired result. Good point, I should have started with this in the first place. I appreciate your patience, @PaigeMiller and @Kurt_Bremser and others. Maybe my choice of procs will turn out to be wrong completely. So, I have raw data, let's say 10 columns, of which one is numeric (the VAR in proc summary). Of this raw data I need to create dozens of partially aggregated datasets, using predefined set of custom formats which I apply to 2-4 columns each time. I need to maintain the structure of the raw dataset, i.e. the 9+1 columns, so I put all 9 in the CLASS. My other idea was that I could have did some remerging after proc summary but I did not like that option, I always try to look for the simplest option. So now I have dozens of aggregated datasets, since they all have the same structure, I wanted to SET them together and create a single dataset, but that failed completely because each dimension adopted the zombie format from the first SET dataset, and as I wrote above the formats differed in each aggregation. Ultimately I wanted to export it into CSV, so I simply exported dozens of CSVs instead of one. This worked. What I meant was that in my situation, it would simply be easier to hardcode formats after each aggregation, perhaps with VVALUE as another member suggested, I thought about that, but I was just hoping there is a neater, cleaner option. If you have any idea or general remarks, I am all ears. Thanks!!
... View more
