Different results by re-arranging data

[berithunsdieck]

I simulated a data example where one biomarker impacts the diagnosis over time directly. If I use a simple lme model as longitudinal model, the effect is not given in the joint model, only after rearranging the longitudinal data by long_data2 <- long_data%>%group_by(SID)%>%slice(1:n())%>%ungroup() the biomarker is significant and the results look completely different. Can you explain why this is the case? Additionally, if the data is not given by long_data2, it is not possible to predict on it (incompatible dimensions).

[drizopoulos]

Can you send a reproducible example?

I simulated a data example where one biomarker impacts the diagnosis over time directly. If I use a simple lme model as longitudinal model, the effect is not given in the joint model, only after rearranging the longitudinal data by long_data2 <- long_data%>%group_by(SID)%>%slice(1:n())%>%ungroup() the biomarker is significant and the results look completely different. Can you explain why this is the case? Additionally, if the data is not given by long_data2, it is not possible to predict on it (incompatible dimensions).

— Reply to this email directly, view it on GitHubhttps://github.com/drizopoulos/JMbayes2/issues/96, or unsubscribehttps://github.com/notifications/unsubscribe-auth/ADE7TTYCD2FO76FRBRGOX3LZHLKYHAVCNFSM6AAAAABJKGUGDCVHI2DSMVQWIX3LMV43ASLTON2WKOZSGM2TGMRQHEYDEMQ. You are receiving this because you are subscribed to this thread.Message ID: @.***>