A*******s
发帖数: 3942 | 1 is there any common SAS option available to "bond" two or more variables
together in forward/backward/stepwise selection? Say if we have many
predictors, we would like two of them stay in model or be dropped out
together. Like Rose said to Jack, you jump I jump. :)
Thanks in advance. |
s*********r
发帖数: 909 | 2 Jack said to Rose?
【在 A*******s 的大作中提到】
: is there any common SAS option available to "bond" two or more variables
: together in forward/backward/stepwise selection? Say if we have many
: predictors, we would like two of them stay in model or be dropped out
: together. Like Rose said to Jack, you jump I jump. :)
: Thanks in advance.
|
A*******s
发帖数: 3942 | 3 get my point! get my point!
【在 s*********r 的大作中提到】
: Jack said to Rose?
|
l***a
发帖数: 12410 | 4 don't think so
【在 A*******s 的大作中提到】
: is there any common SAS option available to "bond" two or more variables
: together in forward/backward/stepwise selection? Say if we have many
: predictors, we would like two of them stay in model or be dropped out
: together. Like Rose said to Jack, you jump I jump. :)
: Thanks in advance.
|
D******n
发帖数: 2836 | 5 write a macro.
【在 A*******s 的大作中提到】
: is there any common SAS option available to "bond" two or more variables
: together in forward/backward/stepwise selection? Say if we have many
: predictors, we would like two of them stay in model or be dropped out
: together. Like Rose said to Jack, you jump I jump. :)
: Thanks in advance.
|
A*******s
发帖数: 3942 | 6 写macro写到吐了...
【在 D******n 的大作中提到】
: write a macro.
|
l*********s
发帖数: 5409 | 7 if you have only 1 group, I think you can do this manually.
【在 A*******s 的大作中提到】
: is there any common SAS option available to "bond" two or more variables
: together in forward/backward/stepwise selection? Say if we have many
: predictors, we would like two of them stay in model or be dropped out
: together. Like Rose said to Jack, you jump I jump. :)
: Thanks in advance.
|
o****o
发帖数: 8077 | 8 也许可以试试EFFECT statement里面的COLLECTION选项
现在好几个working horse PROC都支持EFFECT statement，比如LOGISTIC, PLS,
GLIMMIX, GLMSELECT等
【在 A*******s 的大作中提到】
: is there any common SAS option available to "bond" two or more variables
: together in forward/backward/stepwise selection? Say if we have many
: predictors, we would like two of them stay in model or be dropped out
: together. Like Rose said to Jack, you jump I jump. :)
: Thanks in advance.
|
A*******s
发帖数: 3942 | 9 又学习新东西了，谢谢大牛
包子sent
不过突然有个疑惑，
比如说在logistic里面
我想test H0: beta2=0, beta3=0;
这个collection是直接做一个likelihood ratio test拉倒，还是做俩wald tests with
Bonferroni adjustments ?
【在 o****o 的大作中提到】
: 也许可以试试EFFECT statement里面的COLLECTION选项
: 现在好几个working horse PROC都支持EFFECT statement，比如LOGISTIC, PLS,
: GLIMMIX, GLMSELECT等
|
z**********i
发帖数: 12276 | 10 学习了...
【在 o****o 的大作中提到】
: 也许可以试试EFFECT statement里面的COLLECTION选项
: 现在好几个working horse PROC都支持EFFECT statement，比如LOGISTIC, PLS,
: GLIMMIX, GLMSELECT等
|
d*******y
发帖数: 1154 | 11 膜拜sas帝：）
【在 o****o 的大作中提到】
: 也许可以试试EFFECT statement里面的COLLECTION选项
: 现在好几个working horse PROC都支持EFFECT statement，比如LOGISTIC, PLS,
: GLIMMIX, GLMSELECT等
|
c*******1
发帖数: 342 | 12 likelihood ratio test.
Bonferroni is conservative, right?
with
【在 A*******s 的大作中提到】
: 又学习新东西了，谢谢大牛
: 包子sent
: 不过突然有个疑惑，
: 比如说在logistic里面
: 我想test H0: beta2=0, beta3=0;
: 这个collection是直接做一个likelihood ratio test拉倒，还是做俩wald tests with
: Bonferroni adjustments ?
|
A*******s
发帖数: 3942 | 13 yes LR is less conservative.
But LR requires refitting the model, correct?
Wald statistic with Bonferroni doesn't.
Not sure how SAS implements it.
【在 c*******1 的大作中提到】
: likelihood ratio test.
: Bonferroni is conservative, right?
:
: with
|
