Dear NMusers,
In case that the inter-individual variance (IIV) is not a full block variance
matrix, how to appropriately set up an uninformative Prior for it in NONMEM
when using BAYES method?
I attempted this by trial and error as well as brief reading of NM user guide,
then found that the code might need to be set up as below to get it work. (I
only include the lines of initial estimate and Prior for IIV OMEGA below). But
I am not sure 1) if I should give degree of freedom as 1 for OMEGA associated
with univariate; 2) If an Inverse-Wishart distribution becomes Inverse-Gamma
distribution for IIV OMEGA associated with univariate, then for example, how
does Prior of "$OMEGA BLOCK(1) 0.05 FIX ;CL" play a role? 3) Ultimately, how
to make prior uninformative in this case?
;; Initial estimate
$OMEGA BLOCK(1) 0.4 ; CL
$OMEGA BLOCK(1) 0.4 ; V
$OMEGA BLOCK(1) 0.4 ;KA
;; Prior for OMEGA
$OMEGA BLOCK(1) 0.05 FIX ; CL
$OMEGA BLOCK(1) 0.05 FIX ; V
$OMEGA BLOCK(1) 0.05 FIX ; KA
;;Prior DF for OMEGA
$THETA (1 FIX)
$THETA (1 FIX)
$THETA (1 FIX)
Thanks for your help!
Regards,
Yaming Su