稳健回归 - 浙江大学电子邮件系统.ppt

第八章 稳健回归 §8.1 异常值 我们给出如下的SAS模拟计算的程序： libname tj’d:user’; data tj.a1; do i=1 to 10 ; set tj.p183; r=normal(1234567) ; if i=6 then y=-2-x+12+r; else y=-2-x+r; output; end; run; proc reg data=tj.a1 ; model y=x; output out=b1 p=y1hat; run; data tj.a2; set tj.a1; if x5 then delete; run; proc reg data=tj.a2; model y=x; output out=b2 p=y2hat; run; proc sort data=b1; by x; run; proc sort data=b2; by x; run; data b; merge b1 b2; run; proc gplot data=b; symbol1 v=star c=green I= none; symbol2 v=plus c=red i=l; symbol3 v=plus c=blue i=l; plot y*x haty1*x haty2*x/overlay; run; §7.2 M估计 §7.2 R估计