It is easy to exclude data points in SAS, for example using where statement. The question for you is why those points should be excluded. You need to understand statistics + business first.
可以采用SAS程序的.influence 和cookd距离来进行比较分析，然后进行删除，或者也可以使用**指数VIF.随便放个程序：
proc reg data=jobprof;
/*obtain studentized deleted residuals and hat matrix*/
model y=x1 x3 x4/influence;
/*output Cook distance, DFFITS*/
output out=result1 cookd=cookd dffits=dffits;
ods output outputstatistics=result2;
run;
/*print out cook’s distance values*/
proc print data=result1;
var cookd;
run;
/*obtain t critical value for Bonferroni outlier test*/
data t;
tvalue=tinv(0.999,20);/0.999=1-0.05/(2*25),20=25-4-1*/
run;
proc print data=t;
run;
/*obtain F percentile based on Cook’s distance*/
data result1;
set result1;
percent1=100*probf(cookd,4,21);
run;
proc print data=result1;
var percent1;
run
扫码加好友，拉您进群
全部版块
我的主页
收藏
