COX回归OBSERVATIONS USED个数不同于总观测数？

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

做了一个单因素COX回归，用的MODEL (T1, T2)*CENSOR(0)=VAR/RL=WALD;
其中T1, T2, CENSOR, VAR 都没有缺失值，但是为什么OBSERVATIONS READ是总观测数，而OBSERVATIONS USED差了700呢？
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：observations observation observat tions ATION

Hi,
Because you are using Counting Process Style of Input, i.e., (T1, T2], and those time intervals of T1 >=T2 will be removed from analysis. I think that is the reason why you got fewer observations included in the model.
JingJu

jingju11 发表于 2015-2-6 11:22
Hi,
Because you are using Counting Process Style of Input, i.e., (T1, T2], and those time intervals ...

T1-T2 <= 0啊，应该就是把=0的删除了。谢！

T1-T2=0 was removed as well.

1. data x; set yourdata; ind =(T1<T2); run; proc freq; where not cmiss(of T1, T2, CENSOR, VAR);
   
2. tables ind; run;

复制代码

the count of ind=1 should be equal to # of used obs.

JIngJu

jingju11 发表于 2015-2-6 13:21
T1-T2=0 was removed as well.

还想请教一下，用PROC PHREG如何画分层生存曲线啊？
是不是先生成一个分层数据集，然后在PROC PHREG过程步用BASELINE COVARIATES=stratification dataset就可以了？试验出来好像是可行的。