3 Bayesian analysis
library(PKPDmisc)
library(data.table)
library(tidyverse)
library(knitr)
library(infuser)
Core differences
- Need priors on parameters
- EM algorithms can more robustly handle full block matrices as well as random effects on less well-defined parameters.
3.1 Priors
Priors in nonmem may be defined in two ways:
- By distinguishing the thetas/omegas and their priors in
$PRIOR
- NTHETA=number of Thetas to be estimated
- NETA=number of Etas to be estimated
- (and to be described by NETAxBETA OMEGA matrix)
- NTHP=number of thetas which have a prior
- NETP=number of Omegas with prior
In our case we have 2 Thetas (TVCL, TVV) and 2 omegas (nCl, nV)
$PRIOR NWPRI NTHETA=2 NETA=2, NTHP=2, NETP=2
Or the preferred way of distinguishing in the block definitions
- $THETA - fixed effects
- $OMEGA - random effects
- $SIGMA - residual error
- $THETAP - theta prior
- $THETAPV - variance of theta prior
- $OMEGAP - omega prior
- $OMEGAPD - degrees of freedom to prior omega matrix
** TODO: fill out explanations of each. **