Re: SAS Problem

On Fri, 30 Apr 2004 05:13:05 +1200, Dean Edwards <dean@DREAMNET.CO.NZ>
wrote:

>"Thank you to all the people with your helpful replies to my earlier SAS
problem.

However "SAS Problem" is an extremely vague and uninformative Subject
header.

>
>
>
>My first problem was to get the target points = (number of households for
each area).
>
>and calculate the number of households for 26 areas that earn say $12,824
or less for example. This has been solved.
>
>.
>
>. I have resolved the statistical problem and it is just a coding issue
problem now.

Design and coding are not the same thing. If you leave out design you may
get in some deep stuff.

>
>
>
>Here is my coding problem below:
>
>
>
>The problem below is I have to run this code 26 times separately to
change the target=" " number which changes for each area. Target= "the
number of households."
>
>THE CODING PROBLEM is that I have 26 target ranges to change. = (Area 1
has a target NUMBER=2,313 households for example). In my actual the full
file would be 307,000 households broken into 26 areas multiply by 10
decile target income points = 260.
>
>
>
>
>
>My sample I have 26 households
>
>Area 1 = 13 records My target is 10 housholds for area 1
>
>Area 2 = 13 records My target is 5 households for area2
>
>
>
>This is the only thing I have to change in the program code below.
>
>
>
>Someone suggested a SAS macro. Is this the way to go to get target
automated.

Possibly, but not necessarily. The code below seems rather convoluted, and
I'm not exactly sure of what it is doing. However, I suspect that non-
macro looping and/or grouping can solve the problem. My usual rule is to
avoid macro if there is a good alternative.

>
>I have 26 x 10 = target households to manually change. 260 changes to
make.
>
>I am not a SAS expert. I would call myself barely competent. Anyone who
tells me anything I have to go and get out the SAS manuals to read.
>
>
>
>I am just using a sample of 26 records but have a full file much larger
than this
>
>
>
>data a;
>
> infile "H:\book1.txt" dlm='09'x dsd missover;

Rather than pointing to a file which only you have, why not place the data
inline (with a CARDS statement)? Runnable code is likely to get more
attention.

>
> input area income count Target_Number;
>
>run;
>
>
>
>proc print;
>
>run;
>
>
>
>data b (keep= AREA Target_Number);
>
> set a;
>
>run;
>
>
>
>
>
>data boundary;
>
> set b;
>
> id+1;
>
>run;

The last 3 DATA steps can easily be combined in one.

>
>
>
>* randomly allocate households to the lower decile group
>
> by taking an exact-sized random sample from the decile boundary group;
>
>
>
>data random (keep=readit target decile);
>
> target=10; * must change target number

Here's the immediate problem. There is no semicolon at the end of the last
line. Hence the following line is seen by the compiler as part of the
comment.

>
> if _n_=1 then remains=totobs;
>
>
>
> do while (target>0 and readit<=totobs);
>
> readit+1;
>
>
>
> if ranuni(0) < target/remains then
>
> do;
>
> decile='01';
>
> output;
>
> target=target-1;
>
> end;
>
>
>
> remains=remains-1;
>
> end;
>
> stop;
>
> set boundary point=_n_ nobs=totobs;

So the SET is never executed. The only information taken from BOUNDARY is
the number of observations.

>
>
>
>run;
>
>
>
>* reconstitute dataset with original variables;
>
>data ddd;
>
> set random;
>
> set boundary point=readit;
>
>
>
>
>
>proc print;
>
>run;
>
>
>* followed by another data step to reset the remaining households (not
selected in the random sample) to the higher decile group;

This should be included in the problem statement. Its nature may be
critical to the solution.

>
>
>
>Below is the sample 26 records
>
> Target_
>
> Obs area income count Number
>
>
>
> 1 1 1 1231 0
>
> 2 1 2 1561 0
>
> 3 1 3 1681 0
>
> 4 1 4 2421 10

Is this the same "10" that is to be embeeded in the TARGET= statement? If
so, I don't understand.

>
> 5 1 5 3265 0
>
> 6 1 6 6543 0
>
> 7 1 7 3232 0
>
> 8 1 8 7767 0
>
> 9 1 9 4354 0
>
> 10 1 10 3232 0
>
> 11 1 11 8998 0
>
> 12 1 12 4343 0
>
> 13 1 13 3232 0
>
> 14 2 1 111 0
>
> 15 2 2 232 0
>
> 16 2 3 567 0
>
> 17 2 4 234 0
>
> 18 2 5 666 5
>
> 19 2 6 666 0
>
> 20 2 7 789 0
>
> 21 2 8 698 0
>
> 22 2 9 541 0
>
> 23 2 10 345 0
>
> 24 2 11 344 0
>
> 25 2 12 333 0
>
> 26 2 13 678 0
>
>
>
>
>
>The final output should look something like this below. The program has
been run twice on area 1 and then area2 changing the target number from 10
to 5. I want to run the program once so it will output areas together
>
> Target_
>
> Obs target decile area Number id
>
>
>
> 1 10 01 1 0 1
>
> 2 9 01 1 0 2
>
> 3 8 01 1 10 4
>
> 4 7 01 1 0 5
>
> 5 6 01 1 0 7
>
> 6 5 01 1 0 8
>
> 7 4 01 1 0 9
>
> 8 3 01 1 0 10
>
> 9 2 01 1 0 11
>
> 10 1 01 1 0 13
>
> Target_
>
> Obs target decile area Number id
>
>
>
> 1 5 01 2 0 1
>
> 2 4 01 2 0 2
>
> 3 3 01 2 0 3
>
> 4 2 01 2 5 5
>
> 5 1 01 2 0 6