can two highly skewed variables do correlation?


所有跟贴·加跟贴·新语丝读书论坛

送交者: hopelessless 于 2007-11-23, 01:08:51:

by highly skewed, I mean the variables are not normally distributed (Gaussian distribution). For example, variables are distributed according to power law y=kxa (NOTE: a is superscript of x, both k and a are constants).

I read some statistics, saying that correlation analysis must be done with variables that follow normal distributions. What if two highly skewed variables are indeed correlated each other? I need some related literature about the issue. thanks.




所有跟贴:


加跟贴

笔名: 密码: 注册笔名请按这里

标题:

内容: (BBCode使用说明