→ andrew43: 1 看起來是的 07/12 18:01
→ andrew43: 2 不能說沒意義,就做得比較細而已,或是說目的就是如此 07/12 18:02
也就是說如果自變數間有交互作用,那麼就必須分層去看單純效果;那假如
沒有交互作用,為了研究的目的或興趣也可以分層去看效果,是醬子的意思
嗎?
→ andrew43: 3 可能是因為上標d註解的關係 07/12 18:08
謝謝andrew的回答。只是我實在看不懂d的註解,也不知道這樣會有什麼影響,
統計太差了QQ
※ 編輯: steven225 (122.100.70.103), 07/12/2017 22:11:23
→ andrew43: 是;要看完整文章內容才知道實際模型怎麼建的 07/12 22:31
Linear-mixed models with an unstructured covariance matrix, assuming
independence across participants, were used to model longitudinal differences
in SBP between treatment groups. Fixed effects in the model were BP at
randomization and a treatment group indicator. The time to first occurrence
of the primary composite outcome, all-cause mortality, primary composite
outcome plus all-cause mortality, SAEs, and loss to follow-up or withdrawing
consent were compared between the 2 randomized groups using Cox proportional
hazards regression models with the baseline hazard function stratified by
clinic site (participants were recruited at 100 clinics).
For the primary cardiovascular disease composite outcome, sensitivity
analyses accounting for the competing risk of death were conducted using the
subdistribution hazard model of Fine and Gray.
※ 編輯: steven225 (223.137.131.31), 07/13/2017 11:33:22
※ 編輯: steven225 (223.137.131.31), 07/13/2017 11:35:09
→ andrew43: ... google stratified Cox model 的 stratified 是什麼 07/13 11:46
等比例風險假設(Proportional hazard assumption, PH assumption)是Cox model
的前提假設,我就不知道為何要如此假設。若不符合此假設,必需使用Stratified
※ 編輯: steven225 (223.136.149.11), 07/13/2017 16:44:16