data a;
input class1 $ class2 $ count @@;
cards;
A A 3
A a 1
A b 2
B B 7
B c 1
B d 2
B e 4
;

proc means data=a order=freq nway noprint;
        class class1 class2;
        freq count;
        output out=hell(drop=_type_ rename=(_freq_=count));
run;
proc print data=hell;

最后去除number一列就行了

很正统的处理方法，当大分类的count相同时，有可能会发生混乱，总之非常感谢！！

Hi Deng,

I have meet some problem when use this part of code.

The dataset you can find in attachment are not correctly sorted by proc means.

Maybe I use wrong option or keep wrong observation.

The result show that :

The second observation by Class 1 (the first col) value 117 didn't list in the first place.

So ,please tell me.. Thank you :)

The code I use as this:


10/11 19:12

Maybe I find the problem.

When use proc means, the class class1 both include only class1's sum count and include class1's all sub-count.

So, N(class1) 96 + N(class2 sum) 91 > I(class1) 117 + I(class2 sum) 64

Well, you data flawed, because the # of preferred terms do not add up to the total # given in the body class. In this case, you have to include two frequency variables in the data and PROC SORT it.