c****o
发帖数: 69 | 1 REGRESSION的时候，到底该怎么处理呢？比如有一个变量，只有差，中，好，非常好四
个LEVEL，那我是该用3个DUMMY VARIABLE还是直接当成连续型变量好呢？ | L**********A
发帖数: 7115 | 2 likelihood ratio test
【在 c****o 的大作中提到】
: REGRESSION的时候，到底该怎么处理呢？比如有一个变量，只有差，中，好，非常好四
: 个LEVEL，那我是该用3个DUMMY VARIABLE还是直接当成连续型变量好呢？
| c****o
发帖数: 69 | 3 怎么理解?
你是说两个方法都试，然后比较谁更好？
不过LIKELIHOOD RATIO TEST。。。怎么个用在这上面啊。。。 | j*****e
发帖数: 182 | 4 The best approach is to treat it as ordinal and use generalized linear model
/likelihood method to analyze the data.
However, you need to have a strong stat background to do this and the
interpretation of model parameters is hard for many people.
Furthermore, if the data is collected under a complex design, you have to
apply GEE, which add another layer of complexity.
Thus, in practice, plots of researches just treat it as continuous and
employ general linear model/ANOVA-type approach. | t********y
发帖数: 469 | 5 proportional odds model vs. non-proportional odds model | B******y
发帖数: 9065 | 6 狂赞一个！写的太好了！
model
【在 j*****e 的大作中提到】
: The best approach is to treat it as ordinal and use generalized linear model
: /likelihood method to analyze the data.
: However, you need to have a strong stat background to do this and the
: interpretation of model parameters is hard for many people.
: Furthermore, if the data is collected under a complex design, you have to
: apply GEE, which add another layer of complexity.
: Thus, in practice, plots of researches just treat it as continuous and
: employ general linear model/ANOVA-type approach.
| c****o
发帖数: 69 | 7 Thanks a lot!!
model
【在 j*****e 的大作中提到】
: The best approach is to treat it as ordinal and use generalized linear model
: /likelihood method to analyze the data.
: However, you need to have a strong stat background to do this and the
: interpretation of model parameters is hard for many people.
: Furthermore, if the data is collected under a complex design, you have to
: apply GEE, which add another layer of complexity.
: Thus, in practice, plots of researches just treat it as continuous and
: employ general linear model/ANOVA-type approach.
|
|
