做线性回归遇到的问题

是 否 +2 论坛币

k人 参与回答

经管之家送您一份

应届毕业生专属福利!

求职就业群

赵安豆老师微信：zhaoandou666

经管之家联合CDA

送您一个全额奖学金名额~ !

立即领取

感谢您参与论坛问题回答

经管之家送您两个论坛币！

+2 论坛币

做简单的线性回归：
proc reg data=A outest = B noprint;
     model su = year ;
     by v;
     output out=C;
run;

出现了这样的warning，可是不是太理解为什么：
WARNING: The range of variable YEAR is so small relative to its mean that there may be loss of accuracy in the computations. You may need to rescale the variable to
         have a larger value of RANGE/abs(MEAN), for example, by using PROC STANDARD M=0;
用于回归的记录数是20条。

那位大侠可以帮忙解释一下么。以前也做过这样的回归，用的也是year这个变量，可是当时没有出这个问题，当时的记录数是35条，和这个有关系么？
谢谢！

扫码加我 拉你入群

请注明：姓名-公司-职位

以便审核进群资格，未注明则拒绝

分享0 收藏0 回帖

关键词：线性回归 computations Computation Variable Standard relative example warning

1# redaring

In statistics, variation is information. Same to say, a constant contains no information. Tiny range implies small variation and hence little information. JingJu

2# jingju11

if year range is so small, i do not think you can use linear regression, try logistic regression

redaring 发表于 2010-5-25 10:29
做简单的线性回归：
proc reg data=A outest = B noprint;
     model su = year ;
     by v;
     output out=C;
run;

出现了这样的warning，可是不是太理解为什么：
WARNING: The range of variable YEAR is so small relative to its mean that there may be loss of accuracy in the computations. You may need to rescale the variable to
         have a larger value of RANGE/abs(MEAN), for example, by using PROC STANDARD M=0;
用于回归的记录数是20条。

那位大侠可以帮忙解释一下么。以前也做过这样的回归，用的也是year这个变量，可是当时没有出这个问题，当时的记录数是35条，和这个有关系么？
谢谢！

I believe it should be dependent variable su instead of year. Usually you will be a huge value of intercept. Normalization may help accuracy in the computations.

The message should be,

WARNING: The range of variable su is so small relative to its meanthat there may be loss of accuracy in the computations. You may need torescale the variable to