G*******l
发帖数: 601 | 1 retrospective study, Subjects在吃完一个疗程的药后，停一段时间再吃第2个疗程
的药, 停一段时间再吃第3个疗程的药 (每一个疗程都吃同一总药) 有baseline lab
data，及每个疗程后的lab data 想证明多个疗程是安全的。
请问应该用什么analysis? T-test 似乎不太合适，因为measurement之间可能有
correlation的问题
Thanks a lot! |
m****g
发帖数: 40 | 2 在学校读书的时候遇到过类似问题,考虑用mixed modeling, 同时在确定covariance
structure 的时候 可用 AR（1）
供参考
【在 G*******l 的大作中提到】
: retrospective study, Subjects在吃完一个疗程的药后，停一段时间再吃第2个疗程
: 的药, 停一段时间再吃第3个疗程的药 (每一个疗程都吃同一总药) 有baseline lab
: data，及每个疗程后的lab data 想证明多个疗程是安全的。
: 请问应该用什么analysis? T-test 似乎不太合适，因为measurement之间可能有
: correlation的问题
: Thanks a lot!
|
G*******l
发帖数: 601 | 3 Any other opinion? Thanks |
q******n
发帖数: 272 | 4 You can use repeated measure analysis with one Tx group to test time effect.
There is not control group, right?
Or, you can use growth curve modeling. |
G*******l
发帖数: 601 | 5 qinglian,
No, there is no control group. I was thinking about repeated measure ANOVA,
but the limitation is the correlation issues between measurements. Thanks a
lot! |
q******n
发帖数: 272 | 6 In repeated measure ANOVA, the covariances (correlations) are specified as
homogeneous across time point. It has been taken care of. |
P****D
发帖数: 11146 | 7 just model the correlation structure - the formal term is variance-
covariance matrix.
【在 G*******l 的大作中提到】
: qinglian,
: No, there is no control group. I was thinking about repeated measure ANOVA,
: but the limitation is the correlation issues between measurements. Thanks a
: lot!
|
G*******l
发帖数: 601 | |
