Insert a new variable into dataset, which is the sum total of another variable already in dataset

[H]

As easy as sounds.  Seems like a proc sql, but I would also be interested in see data step option.

data meta;

input A;

datalines;

12

8

14

25

8

16

;

Results

A   B

12    83

8      83

14    83

25    83

8      83

16    83

[H]

This is my way too long code so far...

PROC MEANS NOPRINT DATA=meta;

VAR A;

OUTPUT OUT=summarydata SUM(A) = sigma_A;

DATA meta_summary;

IF _N_=1 THEN SET summarydata;

SET meta;

drop _type_ _freq_;

run;

proc print data=meta_summary;

run;

[H]

PS, I no longer see where to click to edit my posts?

[AnnaBrown]

Hi H,

On the post you want to edit, click on the "..." in the upper right side and select "edit post/reply."

Anna

[Astounding]

Here's a DATA step approach: data results; do until (done1); set meta end=done1; B + A; end; do until (done2); set meta end=done2; output; end; run; Good luck.

[H]

the approach works astounding, but I don't know if it will work for me. In that I will have many more columns to sum and other data steps within this one.

[H]

some reason I thought you could insert the sum back into the dataset within the proc means statement I used.  It would use an "in" I believe.

Any body familiar with that approach?

[Reeza]

PROC SQL solution:

proc sql;
create table B as
select *, sum(a)
from have;
quit;

I don't think you can add the obs back in with proc means, there may be a way with proc summary though, but I'm unfamiliar with that procedure.

[H]

Reeza,

How do you name the new variable in proc sql?  It currently gets named "_TEMG001".

Say I want to call it Sigma_A.

Thanks.

[Reeza]

sum(A) as Sigma_A

[H]

I am using the following to sum two different variables (i.e., A and C):

proc sql;

create table B as

select *, sum(A)as Sigma_A,

sum(C)as Sigma_C

from meta;

drop D;

quit;

But now I would like to sum them based on a group variable in the dataset called replicate. There are 100 replicate groups (i.e., 1-100) all with 300 observations. I would like to execute the about code but have the sums be for the replicates, and inserted into the new dataset as before.

Any help would be appreciated - I am currently having difficulties getting the "group by" to work with the multiple sums in the step.

[H]

I am now using the following, which works out better for my needs (you can see I am looking at 4 variable sums):

proc means data=summies;

class replicate;

var A B C D;

output out=LR_test sum(A)= A_sum

sum(B)= B_sum

sum(C)= C_sum

sum(D)= D_sum;

run;

However, the generated dataset has an extra row for the totals. So it has a replicate  = "." with the totals. Plus the 100 other rows with the sums. Is there a way to get rid of this extra within the above data step?

[data_null__]

Use the PROC statement option NWAY.

[Astounding]

H,

Almost all of what you ask for is relatively easy to program.  But you will need to set a fixed target, not a moving target.  Many replicates?  No problem.  But a different problem.  Name the new fields with "Sigma_"?  No problem.  But a different problem.

One thing you will have to think through is the length of the new variable names.  With an original name like "A", there's no problem creating "Sigma_A".  But what if the original variable name were 30 characters long?  Now there's no room to put "Sigma_" in front.

Anyway, spell out a final form to the problem, and the solution won't be that difficult.