762-763 / 1676
Mixture model
•
Split model in two:
•
A dose-response model as just presented
•
A parametric time to event component
•
Likelihood contributions:
•
Patient experiencing event at time
:
L
+
=p(EUD(n,m),Y
i
)*
f(
)
•
Patient without event until end of follow-up at
•
L
-
=1-p(EUD(n,m),Y
i
)*
F(
)
•
Where F is the cumulative of f