drizopoulos
commented
5 years ago Indeed, in mvJointModelBayes() the association parameter has the name of the longitudinal outcome with the default functional form being the current value, i.e., you should see in the output KTV_value.
With regard to the error in aucJM() check if there are events in this interval and or whether some longitudinal measurements were taken at exactly time 2 or 8. You could add or subtract to them a small quantity, e.g., 1e-06.
Dear Prof. Rizopoulos,
I'm running the following analysis:
long_ktv<-mvglmer(list(KTV~TIME + (TIME|ID)), data=LONG_SCORE, families=list(gaussian)
cox_ktv<-coxph(Surv(DAYS_AT_RISK, DEATH) ~ COUNTRY + AGE + SEX +DIABETES + HEART_COMPOSITE +SMOKING_STATUS, data=SCORE, model=TRUE)
joint_ktv<-mvJointModelBayes(long_ktv, cox_ktv, timeVar="TIME")
I cannot find in the output (summary(joint_ktv)) the "Event process" session and the association parameter alpha (Assoct) as when I run the simple JointModelBayes function. Where can I get this useful information?
Moreover if I try to calculate AUC or the ROC curve I get this error:
aucJM(joint_ktv, LONG_SCORE, Tstart = 2, Thoriz = 8, idVar = "ID")
error: Mat::elem(): index out of bounds Error in log_post_RE_svft(b, Data = Data) : Mat::elem(): index out of bounds
roc_ktv<-rocJM(joint_ktv, LONG_SCORE, Tstart=2, Dt=6, idVar = "ID")
error: Mat::elem(): index out of bounds Error in log_post_RE_svft(b, Data = Data) : Mat::elem(): index out of bounds
What is wrong?
Thank you for your help!