topic Does the Proc tabulate carry the column percent calculation function? in SAS Programming

Does the Proc tabulate carry the column percent calculation function?

ybz12003 — Thu, 05 Jan 2023 15:38:11 GMT

Hello,

I have a sample code below; I would like to calculate the age and ICU column percent of Sum_cost. Is there a way to approach it? Thanks.

proc tabulate data=DataIn;
  var Cost_Sum;
  class age ICU;
  table (age ICU),(Cost_Sum) *(N mean std median p25 p75 sum);
run;

Re: Does the Proc tabulate carry the column percent calculation function?

Cynthia_sas — Thu, 05 Jan 2023 16:15:33 GMT

Hi:
PROC TABULATE has quite a few "percent" statistics: ROWPCTN, ROWPCTSUM, PCTN, PCTSUM, COLPCTN, COLPCTSUM, etc. In addition, with PCTN and PCTSUM, you can ask for a custom denominator. I generally recommend using the built in statistics first to see whether there's one that suits your purposes. The TABULATE documentation has good examples of the percent statistics and how to request a custom denominator: https://go.documentation.sas.com/doc/en/pgmsascdc/9.4_3.5/proc/p0vdza4puzkcghn1j5xtqfb4sahr.htm and this is an excellent paper on the topic: https://support.sas.com/resources/papers/proceedings13/134-2013.pdf
Cynthia

Re: Does the Proc tabulate carry the column percent calculation function?

donricardo — Thu, 05 Jan 2023 17:23:55 GMT

Hi;

While this might be a bit over-simplified, perhaps it'll get you on your way:

data datain;
infile datalines;
input age icu cost_sum;
datalines;
17 1 10000
65 1 13250
44 2 6200
89 2 23000
;

proc format;
picture perc (round) low-high='09.99%' (multiplier=100);
run;

proc tabulate data=datain;
var cost_sum;
class age icu;
table(age icu),(cost_sum) * (n mean std median p25 p75 sum pctsum *f=perc.);
run;

I've used the pctsum keyword, one of Cynthia's suggestions, but often programmers have difficulty getting the needed '%' sign to appear. To do this, you'll need to create your own format as I've done above, using a 'picture' statement instead of the usual value statement. Here are some results using the make-believe dataset above: