e******8
发帖数: 5 | 1 弱弱的问牛人，判断两个变量之间是否correlation时
为什么要使用一下shuffle data？ 然后对比real data 最后才能得出结论
不能通过变量的real data直接得出关联程度吗？
谢谢大牛！！！ | t*d
发帖数: 1290 | | e******8
发帖数: 5 | 3 大牛说的是问题的关键,
通过散点图,我判断应该有弱关联.然后用相关系数分析了这两个变量的两组数据,得到
了弱关联.p值也在合理范围内.
但是我的adviser 说要shuffle其中一个变量,然后再分析下两组数据的关联度,才能确
定是否有关联? 俺是完全的confused了
急求大牛进一步指导!!! 万分感谢!!! | c******d
发帖数: 306 | 4 does ur boss know statistics? sometimes, boss isn't always right.
this question is simple. run a linear regression between these two variable.
if p value less than 0.05, the coefficient is significant, which means
there is a significant correlation between these two variable. but you have
to choose which is x and which is y. since you only have two variables, it
really doesn't matter.
or if you know some programming like R, just use corr(a,b). | t*d
发帖数: 1290 | 5 俺不是科班的，随便说说。
如果你的两组数据符合回归需要的条件，比如参差正态分布，那么就不需要shuffle。
如果不是很肯定，就 shuffle 一下吧。用 shuffle 没有那么多稀奇古怪的要求。
一个小窍门，如果 shuffle 后的pvalue 没有达到显著，但是很靠近了（比如0.
055），就再shuffle 一次。一般多shuffle了几次，总有一次能达到显著的。
【在 e******8 的大作中提到】
: 大牛说的是问题的关键,
: 通过散点图,我判断应该有弱关联.然后用相关系数分析了这两个变量的两组数据,得到
: 了弱关联.p值也在合理范围内.
: 但是我的adviser 说要shuffle其中一个变量,然后再分析下两组数据的关联度,才能确
: 定是否有关联? 俺是完全的confused了
: 急求大牛进一步指导!!! 万分感谢!!!
| e******8
发帖数: 5 | 6 I always think so。 But my boss still want the shuffle result. Actually,i do
not understand why i have to do shuffle...
variable.
have
【在 c******d 的大作中提到】
: does ur boss know statistics? sometimes, boss isn't always right.
: this question is simple. run a linear regression between these two variable.
: if p value less than 0.05, the coefficient is significant, which means
: there is a significant correlation between these two variable. but you have
: to choose which is x and which is y. since you only have two variables, it
: really doesn't matter.
: or if you know some programming like R, just use corr(a,b).
|
|
