More grouping options in proc summary?

[WesBarris]

I have some data that I am analyzing using proc summary.  The following example produces output that is really close to what I need:

data input;

   input source sex $ weight;

   datalines;

176 M 2240

176 F 1814

177 M 2380

177 F 1910

178 M 2190

178 F 1790

179 M 2250

179 F 1850

;

run;

proc summary data=input;

   class source sex;

   var weight;

   output out=summary mean=weight cv=cv n=no;

run;

proc print data=summary noobs;

run;

I would like to group some of the class data together so that the output includes the "176+177" lines shown below:

source	sex	_TYPE_	_FREQ_	weight	cv	no
	.	0	8	2053	11.4592	8
	.	F	1	4	1841	2.8350	4
	.	M	1	4	2265	3.5777	4
176	2	2	2027	14.8608	2
177	2	2	2145	15.4937	2
176+177	2	4	2075	15.1234	4
178	2	2	1990	14.2132	2
179	2	2	2050	13.7972	2
176	F	3	1	1814	.	1
176	M	3	1	2240	.	1
177	F	3	1	1910	.	1
177	M	3	1	2380	.	1
176+177  F	3	1	1860
176+177  M	3	1	2310
178	F	3	1	1790	.	1
178	M	3	1	2190	.	1
179	F	3	1	1850	.	1
179	M	3	1	2250	.	1

The only way I can think of doing that is to use several data steps to combine the values for several of my sources and then append the combined data to my original data before running through proc summary.  Is there a more elegant approach?

[Reeza · Accepted solution]

Use a multilabel format in proc means (rather than summary). I think most of the code will stay the same though between the two procedures.

Base SAS(R) 9.3 Procedures Guide, Second Edition

[data_null__]

MULTILABEL formats work in PROC SUMMARY.  PROC MEANS and PROC SUMMARY are the same program.

[Reeza]

Thanks, are they always interchangeable?

[data_null__]

The defaults are different. 

The documentation for PROC SUMMARY is basically "see proc means"

Base SAS(R) 9.3 Procedures Guide, Second Edition