求教：bootstrap resampling

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

我是用一个cox regression来看一些因素譬如饮酒对骨折的HR，老板建议用bootstrap resampling来做internal validation，不知道怎样用sas？
看网上说都得用macro来写？好复杂，sas新手一个，不会用macro。。
另外请教大牛：proc surveyselect和那些macro有什么不一样的吗？resampling with replacement 和 without replacement，有很大区别吗？
谢谢帮助！！感激！！

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：Resampling Bootstrap Sampling Bootstra boots without 网上

No, macro is unnecessary if you're not going to reuse the code.
Basically you use PROC survyselect to draw samples, then merge in the covariates, and run Cox model with by-group processing to obtain a bootstrapping confidence interval for hazard ratio or whatsoever.
My feeling is that resampling with replacement is preferred if your sample size is close or greater than the # of subjects so at least some samples can avoid the outliers or whatever. I could be wrong.

SAS procedure is simple. The following example will generate 10 sampling sets. You need a text book for what is bootstrap sampling.

proc surveyselect data=sashelp.class out=outboot
seed=123456
method=urs
samprate=1
outhits
rep=10;
run;

proc print;run;

邓贵大 发表于 2013-10-4 09:29
No, macro is unnecessary if you're not going to reuse the code.
Basically you use PROC survyselect  ...

Thanks so much for your help~ really appreciate it!
I tried proc surveyselect with 1000 times resampling, and my pc crashed..
The percentages of missing data was the same high as the original data.. So if I need multiple imputations to handle missing data in the bootstrap sampling, my pc will definitely be dead again..

So can I ask: normally you will deal with missing data before or after the bootstrap resampling?
Big thanks and regards~

bobguy 发表于 2013-10-4 10:37
SAS procedure is simple. The following example will generate 10 sampling sets. You need a text book  ...

Nice reply! Thanks for your patience and help!!