在SAS中要算出有序残差与其期望值的相关系数，应该如何编程？

proc reg data=yourdataname;

model y=x1 x2 x3;

plot r.*p.;

output out=outres p=predicted_y r=residuals;

run;proc corr data=outres;

var predicted_y residuals;

run;

但是光凭肉眼看QQ图上点的分布呈不呈一条直线，以此判断正态性心里不踏实。

You can use proc univariate to test the normality of your variable:

proc univarite data=outres normal;

var residuals;

qqplots residual;

histogram residual;

run;

In your SAS output, if the p value of Shapiro-Wilk or D'Agostino-Pearson tests are p<.05, you can say 凭肉眼看QQ图上点的分布不呈一条直线, that is statistically significant, the 'residuals' is not normally distributed.

热心解答他人疑问，给予100现金奖励！

[此贴子已经被eijuhz于2007-2-22 20:17:06编辑过]

谢谢 不过我是要算nqq.与r.相关系数，用这种方法来初步判定r.是否正态，nqq.如何输出到输出集？？。（再请赐教）

您说的第二种方法很好。

proc univarite data=outres normal;

var residuals;

qqplots residual;

histogram residual;

run;

受教了 谢谢大侠！拜个晚年！！不好意思 过完节了才来逛！！