不同變量的SAS計數 - 數據步驟

Question

我想在沒有proc sql的情況下在數據步驟中執行相同的操作。 注：變量1可變

proc sql; 
select count(distinct(variable1)),variable2,varibale3  
     from tablename group by variable2,variable3; 
quit; 

TIA

Answer 1

一個字符，該解決方案是基於你其實是想你的結果，而無需使用PROC SQL的假設。

打破在不同的步驟解決方案以保持它非常簡單，容易爲你understand.:-

/*Step 1 This step just keeps the required variables in the dataset*/ 

    data tablename(keep=variable1 variable2 variable3); 
    set tablename; 
    run; 

/*Step 2 Sorting and using noduprecs to keep distinct records in the table*/ 
    Proc sort data=tablename nodupkey; 
    by _all_; 
    run; 

/*Step 3 Creating a tag as 1, so as to enable us to sum or count the distinct values*/ 
    Data tablename; 
    set tablename; 
    tag=1; 
    run; 

/*Step 4 Using Proc means to sum the tag inside var on the combination of Variable1 and Variable3 which is inside the Class*/ 

/*I have used PROC MEANS here. You can do the same thing using PROC TABULATE or PROC REPORT*/ 

    proc means data=tablename sum; 
     class variable2 variable3; 
     var tag; 
    run; 

讓我知道，如果你還有任何疑問

Answer 2

我不會用數據步，使用雙處理器頻率。在PROC FREQ

proc freq data=have noprint; 
table variable2*variable3*variable1 /out= first_summary; 
run; 

proc freq data=first_summary noprint; 
table variable2*variable3/out=want; 
run; 

proc print data=want; 
run; 

Answer 3

NLEVELS選項是讓這個最簡單的方法。假設它是按變量2 /變量3排序的，這很簡單。這是一個使用sashelp.class的例子;這裏sex代表變量2和變量3，而age代表計數變量。

proc sort data=sashelp.class out=class; 
by sex; 
run; 

ods output nlevels=levels_age_sex; *nlevels table is output; 
proc freq data=class nlevels;  *nlevels here asked for output; 
by sex;        *grouping variable; 
tables age/noprint;     *do not actually want table; 
run; 
ods output close;     *technically unnecessary but I like symmetry;