a********a 发帖数: 346 | 1 Is there a way to get residual from mixed model just like from a linear
model to have a qq plot to see the normality of the data?
Thanks |
b*****n 发帖数: 685 | 2 just use lm since the residual has nothing to do with random effects |
a********a 发帖数: 346 | 3 Thanks Bullren. I have not heard this before, where do you get this? |
h***i 发帖数: 3844 | 4 草稿纸上推一下,不过需要assumption
【在 a********a 的大作中提到】 : Thanks Bullren. I have not heard this before, where do you get this?
|
b*****n 发帖数: 685 | 5 难道不是这样的吗?residual = y-E[y],E[y]只取决于fixed effects.
【在 a********a 的大作中提到】 : Thanks Bullren. I have not heard this before, where do you get this?
|
a********a 发帖数: 346 | 6 Thanks guys. It sounds correct to me. |
q********i 发帖数: 795 | 7 SAS actually provides two types of residuals, marginal residuals which is Y-
XB_hat, conditional residuals which is Y-XB_hat-ZR_hat. Variance of both
residuals depends on the covariance structure of random effects
【在 b*****n 的大作中提到】 : 难道不是这样的吗?residual = y-E[y],E[y]只取决于fixed effects.
|
h***i 发帖数: 3844 | 8 sas的 glimix如果random effect不是normal,应该 fit的很烂吧。
Y-
【在 q********i 的大作中提到】 : SAS actually provides two types of residuals, marginal residuals which is Y- : XB_hat, conditional residuals which is Y-XB_hat-ZR_hat. Variance of both : residuals depends on the covariance structure of random effects
|