SAS数据集中的累积频率

SAS数据集中的累积频率,sas,cumulative-frequency,Sas,Cumulative Frequency,我的数据集如下所示: Customer Sales 1 15 2 14 3 13 4 11 5 12 6 18 7 21 我需要按销售额百分比对客户进行分类,然后将他们分为高、中、低三类 Customer Sales %Sales 7 21 20% 6 18 17%

我的数据集如下所示:

 Customer Sales
        1    15
        2    14
        3    13
        4    11
        5    12
        6    18
        7    21
我需要按销售额百分比对客户进行分类,然后将他们分为高、中、低三类

 Customer Sales  %Sales
        7    21     20%
        6    18     17%
        1    15     14%
        2    14     13%
        3    13     13%
        5    12     12%
        4    11     11%
铲斗需要基于累积频率:

 Customer Sales %Sales CumFreq Bucket
        7    21    20%     20%   High
        6    18    17%     38% Medium
        1    15    14%     52% Medium
        2    14    13%     65% Medium
        3    13    13%     78%    Low
        5    12    12%     89%    Low
        4    11    11%    100%    Low

因此,正如你所看到的,在前33%的销售额中,任何人都将处于高位,中间33%的销售额将处于中位,而底部33%的销售额将处于低位,因此你需要对列进行求和,对数据集进行排序,然后计算累计百分比。使用自定义格式创建Bucket列

/*untested: don't have access to SAS right now*/

PROC SQL noprint;
/*1st get the total no. of sales and stick it into a macro variable*/
select sum(sales) into: TotalSales
from someCustomertable;

create table topCustomers as 
select 
a.Customer
, a.sales
, a.sales/&TotalSales as salesPerc format=percent11.2
, case
    when calculated salesPerc <=1/3 then "High"
    when calculated salesPerc <=2/3 then "Medium"
    else "Low"
  end as Bucket
from someCustomertable as a
order by 3
QUIT;
data sales;
input  Customer Sales;
datalines;
        1    15
        2    14
        3    13
        4    11
        5    12
        6    18
        7    21
;
run;

proc sort data=sales;
by descending sales ;
run;

proc sql noprint;
select sum(sales) format=best32. into :s from sales;
quit;

proc format;
value pctSales
    0-.33='High'
    .33-.67='Medium'
    .67-1='Low';
run;

data sales;
set sales;
retain total 0;
format pctSales percent8.2;
total = total + sales;
pctSales = total/&s;
bucket = put(pctSales,pctSales.);
drop total;
run;

我知道我们可以从proc-Freq的输出数据集中使用Cumul-Freq,但不确定它是否完全符合我的目的