Thanks Ian,

I thought about sorting the data by the number of unique y in each x, so that, the least number of y group get to pick the y first.

This could be done easily by SQL:

proc sql; select *,count(distinct y) as ny from have group by x order by ny,x,y ;

But then I wasn't sure this is enough, so I didn't post.

On Tue, 18 Dec 2007 18:55:57 +0000, Ian Whitlock <iw1junk@COMCAST.NET> wrote:

>Summary: Difficult problem >#iw-value=1 > >Souga, > >You can be happy with Ya's solution, but I would point out >that > > Obs X Y > ----------------- > 1 A001 30 > 2 A002 50 > 3 A003 60 > 4 A004 40 > >Provides a better solution to Ya's example, since every value of >X has a Y. Consequently Ya's algorithm doesn't always find a >complete solution even when one exists. > >However, for the appropriate sort of Ya's example, his algorithm >does yield a complete solution. > > Obs X Y > ----------------- > 1 A004 40 > 2 A001 10 > 3 A002 50 > 4 A003 60 > >But don't always expect it. > >On the other hand, Ya's solution reduces to Howard's when the file >is properly sorted and and the every pair condition is met. > >Ian Whitlock >============== > >Date: Tue, 18 Dec 2007 12:00:56 -0500 >Reply-To: souga soga <souga1234@GMAIL.COM> >Sender: "SAS(r) Discussion" >From: souga soga <souga1234@GMAIL.COM> >Subject: Re: selecting a unique set of data. >Comments: To: Ya Huang <ya.huang@amylin.com> >In-Reply-To: <200712180439.lBI4WOpB010892@mailgw.cc.uga.edu> >Content-Type: text/plain; charset=ISO-8859-1 > >THANK YOU ALL VERY VERY MUCH, this is exactly what i need. >Just had one question :why is the (100*0) needed in the array statement >:-array yy(1:100) _temporary_ (100*0); > > >On 12/17/07, Ya Huang <ya.huang@amylin.com> wrote: >> >> If NOT every value of X has every value of Y, or number of observation >> in each X varies, one way to get the unique Y is as follows, i.e., pick >> a Y for first x group, record it as used y, then for the next x group, >> check if the y has been used, if so, goes to next y in the group... >> temp array in the code is used to record used y value. >> >> data have; >> X="A001";Y=10;OUTPUT; >> X="A001";Y=30;OUTPUT; >> X="A001";Y=40;OUTPUT; >> X="A003";Y=10;OUTPUT; >> X="A003";Y=60;OUTPUT; >> X="A003";Y=40;OUTPUT; >> X="A003";Y=50;OUTPUT; >> X="A002";Y=10;OUTPUT; >> X="A002";Y=40;OUTPUT; >> X="A002";Y=50;OUTPUT; >> X="A004";Y=60;OUTPUT; >> X="A004";Y=40;OUTPUT; >> run; >> >> data need; >> array yy(1:100) _temporary_ (100*0); >> set have; >> retain found used; >> by x notsorted; >> if first.x then found=0; >> if found=0 and yy(y)=0 then do; >> yy(y)=1; >> used=y; >> found=1; >> end; >> if last.x then do; >> if found=1 then y=used; else y=.; >> output; >> end; >> run; >> >> proc print; >> run; >> >> Obs X Y found used >> >> 1 A001 10 1 10 >> 2 A003 60 1 60 >> 3 A002 40 1 40 >> 4 A004 . 0 40 >> >> Note that for A004, both 60 and 40 have been used, therefore >> assign a missing for it. >> >> >> >> On Mon, 17 Dec 2007 22:51:01 -0500, souga soga <souga1234@GMAIL.COM> >> wrote: >> >> >I apologize for not being specific, anyway what i need is in your first >> >paragraph >> > >> >"In your example every value of X has every value of Y. If this is >> >accurate then sort by X Y, and select first record in first group of >> >X's, second record from the second group, etc. to the last group of >> >X's taking the last record in the group." >> > >> >Unfortunately i do not know how to program this, can someone help! >> > >> >Thanks again, >> >Sa >> > >> >On Dec 17, 2007 10:03 PM, <iw1junk@comcast.net> wrote: >> > >> >> Summary: Specs should precede solution. >> >> #iw-value=1 >> >> >> >> Souga, >> >> >> >> In your example every value of X has every value of Y. If this is >> >> accurate then sort by X Y, and select first record in first group of >> >> X's, second record from the second group, etc. to the last group of >> >> X's taking the last record in the group. >> >> >> >> If your example is not accurate, then it makes no sense for anyone to >> >> give an answer until more is known about the requirements. For >> >> example, suppose there are 5 distinct value of X and 4 distinct values >> >> of Y, then it is impossible to have every value of X in the output >> >> file no matter how the Y values are distributed. Consequently there >> >> is >> >> no solution when all values of X must be used. >> >> >> >> So the real question is - what do you have, and what are the >> >> requirements for what you want? Knowing why would also help us to >> >> know whether the problem is worth thinking about. >> >> >> >> It sounds like some kind of operations research problem. >> >> >> >> Ian Whitlock >> >> ============== >> >> Date: Mon, 17 Dec 2007 19:36:40 -0500 >> >> Reply-To: souga soga <souga1234@GMAIL.COM> >> >> Sender: "SAS(r) Discussion" >> >> From: souga soga <souga1234@GMAIL.COM> >> >> Subject: selecting a unique set of data. >> >> Content-Type: text/plain; charset=ISO-8859-1 >> >> >> >> Hi, >> >> >> >> X="A001";Y=10;OUTPUT; >> >> X="A001";Y=20;OUTPUT; >> >> X="A002";Y=10;OUTPUT; >> >> X="A002";Y=20;OUTPUT; >> >> >> >> I need the output to be: >> >> >> >> X="A001";Y=10;OUTPUT; >> >> X="A002";Y=20;OUTPUT; >> >> >> >> ie. the value of y should not repeat for x >> >> >> >> Thanks, Sa >> >> >> >> >>