Hi,

I=92ve a dataset where each row includes a unique provider id and 10 $
variables equal to the sum of charges for a specific procedure code
(CPT code). The 10 procedure codes variables are named with the
following convention:

P_5digitsCPT

So, I might have p_90806 and p_40890 and so on. The program id data
driven so I could end up each time with 10 different codes.

Since all of these start with p_ I can easily define and array using
the p_: convention.
Now, I want to define 10 new variables representing percentages for
each p_ variable (i.e. percent_90806=3Dp_90806/SUM(of p_:).
The new 10 percentage variables should be named percent_5digitCPT.
If I could create a second array for the percent_ variables then the
calculation would be easy.
But how do I create my percent_ variables to reflect the CPT reported
in the original 10 p_ variables? I need to extract the CPT code from
each p_ variable, then create a new variable named using percent_ as a
prefix and then the CPT code.
How do I do that?

Thanks,

G

On Feb 23, 2:20=A0pm, adjgiulio <adjm...@comcast.net> wrote:
> Hi,
>
> I=92ve a dataset where each row includes a unique provider id and 10 $
> variables equal to the sum of charges for a specific procedure code
> (CPT code). The 10 procedure codes variables are named with the
> following convention:
>
> P_5digitsCPT
>
> So, I might have p_90806 and p_40890 and so on. The program id data
> driven so I could end up each time with 10 different codes.
>
> Since all of these start with p_ I can easily define and array using
> the p_: convention.
> Now, I want to define 10 new variables representing percentages for
> each p_ variable (i.e. percent_90806=3Dp_90806/SUM(of p_:).
> The new 10 percentage variables should be named percent_5digitCPT.
> If I could create a second array for the percent_ variables then the
> calculation would be easy.
> But how do I create my percent_ variables to reflect the CPT reported
> in the original 10 p_ variables? I need to extract the CPT code from
> each p_ variable, then create a new variable named using percent_ as a
> prefix and then the CPT code.
> How do I do that?
>
> Thanks,
>
> G

Variable names are compile time thingy, so you can directly
dynamically create new var based on old var.
One way to do it is to get a list of variable using meta data.

Here is another way, use proc transpose:

data have;
array p(5) p_12354 p_23412 p_45323 p_5663 p_32334 (1 3 9 20 8);
do id=3D1 to 2;
output;
end;
run;

data want0;
 set have;
length _name_ $30;
array p(5) p_:;
do i=3D1 to 5;
 _name_=3Dvname(p(i));
 val=3Dp(i);
 output;
 _name_=3D'pct_'||substr(vname(p(i)),3);
 val=3Dp(i)/sum(of p_:);
 output;
end;
keep val _name_ id;
run;

proc transpose data=3Dwant0 out=3Dwant (drop=3D_name_);
var val;
id _name_;
by id;
run;

proc print;
run;

             pct_             pct_
pct_                            pct_
id p_12354   12354  p_23412   23412  p_45323  45323  p_5663 pct_5663
p_32334  32334

 1    1    0.024390    3    0.073171    9    0.21951   20
0.48780    8    0.19512
 2    1    0.024390    3    0.073171    9    0.21951   20
0.48780    8    0.19512

HTH

Ya

try Proc Contents
u can use fact that convention o naming variables is strict and
perform some operations on proc contents output to create macro
variables(those will contain statements which represent calculations
(percent_90806=p_90806/SUM(of p_:). )

either create n macro variables and then perform loop or create one
marco variable in macro with loop in it (marco variable should grow in
each step by content of ith calculation)

proc contents is cool:)
i assume that each executing of code which precede mention set implies
fact that variables names once can be
p_90806 and once p_666 you can easily parse variables names to create
new name like percent_90806 or percent_666

using proc contents with some cool macro desribed above will allow u
to be independent from numbers in those names

regards
tomek