topic Re: combine several output to form a table in SAS Programming

combine several output to form a table

fengyuwuzu — Mon, 29 Aug 2016 20:50:09 GMT

I run the same codes for different data sets (run 5 times the same macro by changing data set names) and got the following output. Is there a way to output into a table, not 5 separate ones like this?

Obs	n	mean	std	median	P50	P55	P60	P65	P70	P75	P80	P85	P90	P95	P98	P99
1	8070	124.494	69.8303	103	103	112	121	133	147	162	182	204	232	274	309	329

Obs	n	mean	std	median	P50	P55	P60	P65	P70	P75	P80	P85	P90	P95	P98	P99
1	7264	132.550	69.0437	112	112	120	130	142	156	172	190	212	239	278	313	330

Obs	n	mean	std	median	P50	P55	P60	P65	P70	P75	P80	P85	P90	P95	P98	P99
1	6404	142.296	67.8522	122	122	131	141	154	167	183	199	220	245	283	316	332

Obs	n	mean	std	median	P50	P55	P60	P65	P70	P75	P80	P85	P90	P95	P98	P99
1	5605	152.690	66.2807	133	133	143	154	165	179	193	210	229	253	290	321	334

Obs	n	mean	std	median	P50	P55	P60	P65	P70	P75	P80	P85	P90	P95	P98	P99
1	4782	165.087	64.0314	148	148	157	167	179	191	205	220	239	263	297	325	337

Re: combine several output to form a table

Reeza — Mon, 29 Aug 2016 21:04:21 GMT

You can append tables fairly easily, just make sure to give them names that start with the same prefix that isn't used elsewhere.

data want;

set table1-table5;

run;

Re: combine several output to form a table

ballardw — Mon, 29 Aug 2016 21:10:18 GMT

Options abound for something like this. But first I would think about providing identification as to where each row of the data in the combined dataset comes from and which variable was summarized.

If the same variable is being summarized from each data set I would be strongly tempted to combine the data sets and use the INDSNAME option on a set statement to add a variable with the source the dataset. Then use that varaible in a Class or By statement in Proc Means/Summary or Univariate.

If the VARIABLE is different then it may be easier to create output datasets and append together but again I suggest having information about either the data source, variable or both in the result.

We would need to see what your macro code is to make suggestions as to which may be preferable.

Re: combine several output to form a table

fengyuwuzu — Tue, 30 Aug 2016 15:56:52 GMT

/* I have 5 variables, and for each one, I want to get the top 6%, 7%, 8%, 9%, 10% data */
%macro percent(var=);
proc univariate data=percent.simple ;
var &var ;
output out=percent pctlpre=P pctlpts= 90 91 92 93 94;
run;

data percent.&var._p90 percent.&var._p91 percent.&var._p92 percent.&var._p93 percent.&var._p94;
if _n_=1 then set percent;
set percent.simple;
p90_flag = &var ge P90;
p91_flag = &var ge P91;
p92_flag = &var ge P92;
p93_flag = &var ge P93;
p94_flag = &var ge P94;
if p90_flag=1 then output percent.&var._p90;
if p91_flag=1 then output percent.&var._p91;
if p92_flag=1 then output percent.&var._p92;
if p93_flag=1 then output percent.&var._p93;
if p94_flag=1 then output percent.&var._p94;
drop P9: ;
run;
%mend percent;

%percent(var=betting_days)  /* run this macro using one variable to generate 5 data sets containing top 6% - 10% data */

%macro print(var=,p=);
proc univariate data=percent.&var._&p noprint;
var &var ;
output out=tmp mean=mean median=median n=n std=std pctlpre=P pctlpts= 60 70 75 80 90 95 98 99;
run;
proc print data=tmp; 
title "&p &var" ;
run;
%mend print;

%print(var=betting_days,p=p90)
%print(var=betting_days,p=p91)
%print(var=betting_days,p=p92)
%print(var=betting_days,p=p93)
%print(var=betting_days,p=p94)

Above are my macros.