I can see why someone might prefer it that way. I think it is in fact one of the options when installing SAS. I personally don't prefer it. I like being able to double-click and open into the SAS environment. One advantage is that that sets the folder in which the .sas file is stored as the work directory. ... View more

RobH SAS Employee' solution is the correct answer.  It requires admin access, which I do not have, but it outlines steps that my organization IT's is able and willing to follow.  Thanks all! ... View more

Freelance Reinh has come closest to a solution so far.  The SAS Global Forum 2013 paper referenced describes the underlying mechanics in great detail but the solutions it offers are now obsolete. Here is what I have tried, following the paper.  1. The SAS file Type Management no longer exists. It would be the best solution, so if anyone knows of an equivalent, that would be very good. 2. Editing the registry directly is not possible since I work in an organization that does not give me admin access to my computer. In addition, our IT support can do simple troubleshooting, but will not touch the registry. 3. I associated .sas files with SASOACT.exe, which is the least preferred solution in the paper, but is the only one available to me. This correctly launches the SAS environment when double-clicking a .sas file, but surprisingly the program does not open in the editor.  ... View more

Yes, you are right, I was sloppy in creating the datasets Have and Want by hand. I like your solution a lot and will be keeping it. So the thread contains three excellent and very different solutions. ... View more

It is simple, but the Want dataset has only one observation. This is what is expected for the two-way combinations only: data want; input set1 $ set2 $ shared_count; datalines; var1 var2 4 var1 var3 9 var1 var4 8 var2 var3 3 var2 var4 2 var3 var4 5 ; run; ... View more

Additional notes. Thank you to all. The community's generosity is always amazing. KSharp answered the original question exactly and his solution is very economical. It requires understanding IML. Tom also answered the question exactly, and his solution is easier to understand and modify for someone with my current skill level. Reeza's suggestion to use PROC DISTANCE does provide a great solution to the problem that justified my question. In fact, one would use PROC DISTANCE to compute the Jaccard index, the PROC CLUSTER to inform how to group tasks optimally.  In other words Reeza's suggestion leads to a better solution than the one that justified my question. That said, I learned several new methods and tips that I am saving to my set of favorite snippets. ... View more

Thank you! I was struggling with how to define the elements in  sum(element(a,b) & element(a,c)); for the higher order combinations. Your additional examples for combinations of four and five  variables demonstrate how to do it. ... View more

KSharp, this looks extremely economical, and of course it works perfectly as requested. I am among the users who have never used PROC IML. How would you extend this to more than three-way combinations? ... View more

Reeza, by "this form, do you mean the tall form? PROC DISTANCE, then CLUSTER might in fact be a good solution to the underlying question (help users decide which vars i.e. tasks can be combined so as to minimize duplication of effort) effort) ... View more

@Tom wrote: So you have 40 lists of IDS. Might be easier if made it into a single list with two variables. Yes, that is a much better way to describe the data. I like the simplicity of this approach a lot. Two questions: 1. In the final proc sql step, I would have to spell out all the combinations I want. Suggestions on how to generate all pairwise combinations? All triplets, etc... 2. Unfortunately, my variables are not named as conveniently as var1-varn. Suggestions on handling that as economically as possible? 3. Likewise, my IDs are numerical, but not as friendly as 1-30. Should I pull the min and max as macro variables and use them in the array?  ... View more

> You seem to be treating this data as 4 independent one variable datasets < In a way, yes, that's correct. See post. The observations don't really "matter". All I want to know is how many points are in the intersection of any two, three, or four of those sets. Thus in the one observation you show, the intersection of var 1 and var2 is empty, but the intersection of var1 and var4 contains 1 point (whose value happens to be 1, so that's confusing) ... View more

This seems to work as expected, and fast. I am glad to see that it is a lot less simple than what I had been trying. Next I need to figure out how to generate triplets. Your suggestions are appreciated !   ... View more

Hope I am not getting them wrong again. I am just doing a manual count. the shared values between var1 and var2 are 5, 6, 25, thus count=3 ... View more

Apologies!  As Tom noticed, some of my counts are incorrect. I mentioned in the post that there are no duplicates within each variable, and my "have" data had some. Here is the corrected data. As Tom pointed out, PROC SUMMARY gets overwhelmed when I try it with the full datasets. data have; input var1 var2 var3 var4; datalines; 1 5 2 1 2 6 3 2 5 19 6 4 6 20 10 6 7 25 13 7 8 26 14 8 12 28 17 12 15 . 21 13 17 . 22 28 21 . 23 30 22 . 25 . 23 . 27 . 25 . 28 . 29 . 30 . 30 . . . ; run; data want; input set1 $ set2 $ shared_count; datalines; var1 var2 3 var1 var3 8 var1 var4 7 var2 var3 3 var2 var4 2 var3 var4 5 ; run; ... View more