程序求解

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

data dd;
input f e n d@@;
lgn=log(n);
cards;
1 1 240 216  1  0  100  63  2  1  60  48  2  0  210  84
;
proc genmod data=dd  order=data;
class f;
model d=e f/dist=poisson
link=log   type1 type3;
run;
proc genmod data=dd;
class f;
make 'obstats' out=outdata;
model d=f/dist=poisson link=log offset=lgn
type1 type3  obstats;
run;
proc print data=outdata;
run;
data k; set outdata;
set dd;
rate=exp(xbeta);run;proc print;
run;

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：poisson genmod OBSTAT Offset model 程序

It is a typical log-linear model fitted in generalized linear regression models in SAS. you are measuring the rate or proportion whe you are using offset variable.
只是你应该注意的是泊松分布是小概率事件。216 out of 240? Unless they are in different units.比如说，240万辆车保险里每月216个重大事故。
jingju

jingju11 发表于 2015-1-19 10:26
It is a typical log-linear model fitted in generalized linear regression models in SAS. you are meas ...

前面那两个1 1 是啥

看不明白你的问题

llkknnllyytt 发表于 2015-1-19 10:29
前面那两个1 1 是啥

f=1, 2; e = 1, 0.
They could be factors that have impact on the counts.
Class f is to take f as a classification variable in the model. Say, f = 1 vs. 2 can be male vs. female, older vs. younger,…
e is treated as continuous in the model;  But for the dummy variable 0/1, they are no difference if you treat it as continuous or discrete.

JingJu

jingju11 发表于 2015-1-19 10:37
f=1, 2; e = 1, 0.
They could be factors that have impact on the counts.
Class f is to take f as  ...

请问可以去掉一个1吗

你可以把它像成是二联表格(2 by 2)的形式。e 和 f 把数据自然分割成四个cells.因而有四个均值。比如

高胖(11)；高瘦(12)；矮胖(01)；矮瘦(02)。如果只有高矮，只能告诉高度的影响；胖瘦只能告诉你体重的影响。而两个变量可以告诉身高和体重组合的影响。

京剧

jingju11 发表于 2015-1-19 11:02
你可以把它像成是二联表格(2 by 2)的形式。e 和 f 把数据自然分割成四个cells.因而有四个均值。比如

高胖 ...

请问你会看poisson程序结果吗AIC值啥意思