Date: Fri, 21 Dec 2007 15:29:12 0600
ReplyTo: "data _null_," <datanull@GMAIL.COM>
Sender: "SAS(r) Discussion" <SASL@LISTSERV.UGA.EDU>
From: "data _null_," <datanull@GMAIL.COM>
Subject: Re: Running deductions based on by variables?
InReplyTo: <44aebe13fd3247259d4fc1c735c4662c@d4g2000prg.googlegroups.com>
ContentType: text/plain; charset=ISO88591
data work.test;
input x y max @@;
cards;
1234 02 3 1234 01 3 1234 00 3
1235 03 5 1235 03 5 1235 01 5
1236 02 3 1236 02 3
;;;;
run;
data work.test;
do until(last.x);
set work.test;
by x;
if first.x then do;
s = sign(max);
r = abs(max);
end;
a = min(r,abs(y));
r = r  a;
a = a * s;
output;
end;
drop s r;
run;
proc print;
run;
Obs x y max a
1 1234 2 3 2
2 1234 1 3 1
3 1234 0 3 0
4 1235 3 5 3
5 1235 3 5 2
6 1235 1 5 0
7 1236 2 3 2
8 1236 2 3 1
On Dec 21, 2007 3:02 PM, Floyd Moseby <floydmoseby@gmail.com> wrote:
> Folks, I am having a tough time with this one. Here's the basic
> scenario: I have a table with observations sorted by two variables,
> the first ascending (X) and the second descending (Y). Each by group
> (designated by variable X) must be allocated a finite number of
> widgets. The widgets will be allocated to the observations according
> to variable Y; the number of widgets allocated to each observation are
> determined by the value of variable Y, so that observations with the
> highest Y value are allocated the most widgets, and the widgets are
> allocated to those observations first. Most by groups will not be
> able to allocate the widgets to more than the first couple of
> observations; once the number of widgets have been allocated in full,
> the remaining observations of that by group get 0.
>
> Here's some sample data:
>
> var X var Y max widgets
> 1234 02 3
> 1234 01 3
> 1234 00 3
> 1235 03 5
> 1235 03 5
> 1235 01 5
> 1236 02 3
> 1236 02 3
>
> What I need to end up with would resemble this:
>
> var X var Y max widgets allocated widgets
> 1234 02 3 2
> 1234 01 3 1
> 1234 00 3 0
> 1235 03 5 3
> 1235 03 5 2
> 1235 01 5 0
> 1236 02 3 2
> 1236 02 3 1
>
> Intuitively, I would use lag variables with conditional statements,
> but it's clear that lag variables do not work within conditional
> statements. Preceding the conditional logic with lag variable
> creation comes close, but the data still doesn't yield the expected
> results.
>
> I'm sure I will kick myself once I figure it out, but hopefully you
> guys can help me kick myself sooner rather than later. Any help?
>
> floyd
>
