How to do nested counts

[cluelesssas]

I'm a beginner to SAS. I'm trying to have nested counts presented cleanly as follows.

This is code to generate toy data, as well as 3 attempts I tried (from hours of googling how to do this) that give the correct info, but none are as clean as the above. Imagine how awful they'd look when I need more than 2 layers of nesting. and multiplying together the total number of categories per variables is large, but the actual number of their combinations in the data is small.

data toy;
	input A $ B $;
	datalines;
A1 B1
A1 B2
A1 B2
A2 B1
A2 B1
A2 B1
A2 B2
A2 B2
A2 B2
A2 B2
;

proc tabulate data=toy;
	class A B;
	table A*(B all) all;
run;
proc summary data=toy print;
	class A B;
run;
proc freq data=toy;
	tables A*B / nopercent norow nocol;
run;

[Cynthia_sas · Accepted solution]

Hi,

My tendency would be to use PROC REPORT. I think report #3 in my example (using your test data) is closer to what you want. You can generate the PROC  REPORT output without any column headers at all; however, I rarely find a need to do that and using column headers keeps the report columns less confusing.

Cynthia

[cluelesssas]

thx Cynthia. What I ended up doing was based on your #2 proposal. I actually had 3 layers of nesting in my actual data (so imagine a variable C that we also want to nest under A and B in the toy data). This code seemed to do the job I hope it will generalize to future situations

proc report data=toy;
column A B C n;
define A / group;
define B / group;
define C / group;
define n / 'Count';
break before A / summarize;
break before B / summarize;
run;

[Cynthia_sas]

Hi:
That is a simpler form of PROC REPORT since it doesn't have any compute blocks or style overrides. If that works for you, then that is going to be easier to code. I suggest you look on www.lexjansen.com for PROC REPORT papers if you are new to using PROC REPORT.
Cynthia

[Ksharp]

data toy;
	input A $ B $;
	datalines;
A1 B1
A1 B2
A1 B2
A2 B1
A2 B1
A2 B1
A2 B2
A2 B2
A2 B2
A2 B2
;

proc freq data=toy noprint;
	tables A*B /out=have list nopercent norow nocol;
run;
proc report data=have nowd ;
columns a b count;
define a/group noprint;
define b/group 'level';
define count/analysis 'count';
compute before a;
b=a;
endcomp;
compute b;
if missing(_BREAK_) then call define(_col_,'style','style={indent=0.25in}');
endcomp;
break before a/summarize;
run;

[SteveDenham]

I know this has been answered, but wouldn't a PROC MEANS (or SUMMARY) give all of the levels needed for this? It would then just be a sorting task to get something to print out. I have to admit the PROC REPORT approach more quickly yields a more esthetically pleasing output, but if the dataset is really large, with a lot of nesting, and you need to use the results for additional work, it might be worth taking a look.

SteveDenham

[Ksharp]

Yes.
But OP need pretty good layout like some margin before some levels.
That is unable to be done by PROC MEANS.
Since OP just only want a REPORT not a dataset, I think PROC REPORT is the only choice.