背景：有時(shí)我們要觀察各個(gè)分?jǐn)?shù)區(qū)間的用戶，在各個(gè)特征上的表現(xiàn)有無差異。在進(jìn)行分組時(shí)，除了使用PROC FORMAT手工定義區(qū)間之外，也可以使用PROC RANK和PROC FORMAT，利用分?jǐn)?shù)（或者其他數(shù)據(jù)）的分位數(shù)等統(tǒng)計(jì)量，實(shí)現(xiàn)自動(dòng)化分組排序。

PROC RANK

proc rank data=test out=r_test【輸出的數(shù)據(jù)集】;

var spend【對(duì)spend進(jìn)行排序】;

ranks r_spend【序號(hào)變量命名為r_spend】; ?

run;

PROC UNIVARIATE

proc univariate data=events noprint;

var neg_score;

output out=p pctlpre=P_【分位數(shù)變量名稱的前綴為P_】

pctlpts=10 to 100 by 10;

weight SamplingWeight;

run;

proc transpose data=p out=pt;

run;

proc sort data=pt

nodupkey force noequals;

by COL1;

run;

Generating deciles, quartiles, percentiles or other groups from numeric variables. The GROUPS optionis used here to specify the binning. Deciles are created by specifying GROUPS=10, quartiles can be generated by GROUPS=4, and percentiles are created with setting GROUPS=100.