Model syntax

As shown in the various examples, models are generated using new_stan_model(). The main requirement is that the ODE model dAdt is defined.

The variable names in the table below are reserved, i.e. they cannot be used in the ODE model code, since PKPDposterior adds these to each model so that it is possible to perform structured analyses.

The following variables are recognized, with their approximate equivalent in NONMEM, if available.

Variable name	Meaning	NONMEM equivalent
dv	Vector of observed values	DV column in input
ipred	Vector of predicted values for all event times	IPRED column in output
ipred_obs	Vector of predicted values for observation records only	IPRED column in output for records where EVID==0
ipred_ruv	Vector of predicted values for observation records only, with residual variability simulated	Y in model file, DV in output table
n_t	Number of event records	Number of rows in input dataset
n_cmt	Number of compartments
n_obs	Number of observations	Number of rows in input dataset with EVID==0
i_obs	Index of observation records within all event records
theta	Vector of model parameters	$THETA (although not same data format)
ode	Function for the ODE system	$ODE
A	Vector of amounts in compartments	A in model file
dAdt	Vector of gradients for ODE system	DADT in model file
prior_<parameter>	Samples of prior parameter distribution

For models with more than one observation type, such as PK-PD models, or a parent-metabolite model, the variables related to observations (i.e. the variables listed in the first 8 rows) are appended with a tag for the respective observation type, e.g. dv_pk and dv_pd, or dv_parent and dv_metab, etc. See the neutropenia model for an example.

If you want to leverage functionality within PKPDposterior, but you cannot (or prefer not to) use new_stan_model() to create the model, you should use the above variable names as well, so that output can be parsed appropriately by PKPDposterior functions.