[NMusers] PK/PD modeling - fixing population paramterers vs. fixing both pop. and individual parameters

[nonmem]

Hello NONMEM Team,
We usually try to separate PK and PD models and the reasons are obvious. There 
are different ways to separate the models.  Are there any good references which 
show advantages and disadvantages of fixing population paramterers in the PD 
model ann allowing the model to estimate individual PK parameters vs. fixing 
both population and individual PK parameters?  
Thanks!
Pavel