Structural_parameters_and_group

Structural parameters and group statements

There are two levels of a model:

The structural level, which has compartments, doses, and observations, and parameters (called structural parameters)

The population level, which has typical values of parameters (fixed effects), inter-individual variation (random effects), covariates, and covariate effects (which are also fixed effects)

The statement that ties the two levels together is called stparm. It defines each structural parameter as a function of fixed effects, covariates, and random effects for estimation and simulation. The model can include any number of stparm statements. The stparm statements are only re-executed during a given iteration if a covariate changes. Hence, they should not include variables that need to be evaluated all the time.

Sometimes, to simplify the stparm statements, it is desirable to calculate parts of them outside in separate statements. This cannot be done with ordinary assignment statements like A=B+C but it can be done inside a group statement, like group(A=B+C), where B and C can only be functions of covariates and fixed effects. The group statement ensures that the block assignment statements get executed prior to the stparm statements.

The parameter defined by the group statement is called a group parameter and it can be used in the right-hand-side of an stparm statement. In situations with complex covariate effects, this can lead to substantial performance improvement, by avoiding frequent calls to math functions.

Example 1:

A covariate model can be defined in using standard PML, as follows:

   stparm(V=((Gender==0) ? tvV:  tvV *dVdGender)*exp(nV))

Each group can be defined separately (by combining covariate values):

   group(
     tvVMale=tvV
     tvVFemale=tvV *dVdGender
     stparm(V=((Gender==0) ? tvMale : tvFemale)*exp(nV))
   )

Example 2:

The following calculates body surface area (BSA) in a group statement using covariates weight (WT) and height (HT), and is only done when covariates change, not on every ODE evaluation. This group statement avoids the calculation of BSA at each iteration of ODE solver steps.

   covariate(WT) # WT: body weight
   covariate(HT) # HT: height
   # BSA: body surface area
   group(BSA=(WT^0.5378)*(HT^0.3964)*0.024265)