PMXStanModel: Creation of a 'PMXStanModel' object

In stanette: R Interface to Stan

Creation of a PMXStanModel object

Description

Initializes an object of class PMXStanModel with methods for generating and compiling Stan code and querying model specifications.

Usage

PMXStanModel(type = "PK", path = "model_temp", route = "1st_order_abs",
  solver = NULL, ode = NULL, pk.struct = "2-cmpt", pk.param = "CL_V",
  obs.state = NULL, theta = NULL, eta = NULL, fixed = NULL,
  compile = FALSE)

Arguments

type	a string to specify the type of model: "PK" (default) or "PKPD".
path	a string to specify the path that will be used to store the model, Stan input data, and model fitting and diagnostic results.
route	a string to specify mode of drug administration: "1st_order_abs" (default), "IV_bolus", or "IV_infusion".
solver	a string to specify which solver to be used for a PK model: "closed_form" (default) or "ODE"; ignored for PKPD models in ODE form.
ode	a string to specify equations of the ODE system for a PKPD model; ignored for PK models.
pk.struct	a string to specify the PK model structure: "1-cmpt", "2-cmpt" (default), or "3-cmpt"; ignored for PKPD models in ODE form.
pk.param	a string to specify the method for PK model parameterization: "CL_V" (default), or "micro_rate"; ignored for PKPD models in ODE form.
obs.state	an integer to specify the index of the state variable corresponding to observed data for a PKPD model in ODE form; ignored for PK model with closed-form solution, but required for PKPD models.
theta	a string or vector of strings to specify which parameters will be estimated for a PKPD model in ODE form; ignored for PK model with closed-form solution, but cannot be NULL for PKPD models.
eta	a string or vector of strings to specify which parameters will have inter-individual variability for a PKPD model in ODE form; ignored for PK model with closed-form solution, and can be NULL for PKPD models.
fixed	a vector of strings to specify which parameter values will be fixed at constants for a PKPD model in ODE form; ignored for PK model with closed-form solution, and can be NULL for PKPD models.
compile	a logical variable indicating whether to compile the generated Stan code during the initialization process (TRUE) or not (FALSE, as default). Note that a PMXStanModel object can be compiled at any time after initialization by calling the compile method (see Value and Details sections).

Details

PMXStanModel serves as an interface for practical PK/PD modeling using Stan under a Bayesian framework. The first step of building such a model is for a user to provide model specifications (for more details on specification arguments and default values, please refer to the Arguments section). With a proper set of specifications, a model-specific Stan source code is then generated based on a generic template code for the associated model type. Users then can choose to compile the auto-generated Stan source code directly to a self-contained platform-specific executable, or to modify the Stan source code according to their own modeling strategies before compiling it to an executable.

The template Stan code serves at least two purposes. First, it presents a grammar-corrected ready-to-be-used code that already takes care of most technical details necessary to build such a model in Stan; secondly, it makes sure that everything needed for running Stan sampling is internally consistent, from preparing the compatible data list, interpreting various dosing events and schedule, to calling the appropriate solver. In a template Stan code, model parameters to be estimated are defined as theta's and inter-individual variabilities for parameters are defined as eta's, to follow the conventions in pharmacometrics practice. All priors are set as non-informative. It needs to be kept in mind that the code is fully accessible and modifiable by a user; therefore, a user can add his/her own customized code by changing certain part of the auto-generated code conveniently. The modified Stan code can be re-compiled at any time as long as the changes made by users comply with standard Stan grammar. It is suggested that in these cases, a user begins with making small changes and testing the compilation (as well as compatibility with the input data if necessary) gradually before going to more significant modifications.

Value

A PMXStanModel object, with the following list of methods:

get.model.specs	returns model specification.
generate.stancode	generates Stan code according to model specification.
get.state.var	returns names of state variables in the ODE system.
get.ode.par	returns names of parameters in the ODE system.
get.ntheta	returns number of parameters to be estimated for an ODE-based PKPD model.
get.neta	returns number of inter-individual variabilities set by user for an ODE-based PKPD model.
get.stan.file	returns path of the auto-generated Stan file.
compile.stanmodel	compiles the Stan file, either auto-generated (default) or with a path specified by user. This function has a generic form compile.
get.compile.status	returns a logical indicator whether the model-associated Stan file has been compiled or not.
retrieve.stanmodel	returns the compiled Stan model.
print.model	returns the compiling status and the path of the model-associated Stan file, as well as model specifications. This function has a generic form print.

Author(s)

Yuan Xiong and Wenping Wang

References

The Stan Development Team. Stan Modeling Language User's Guide and Reference Manual. https://mc-stan.org.

See Also

prepareInputData for transformation of a NONMEM-readable dataset to a list compatiable to auto-generated model-specific Stan code; PMXStanFit for how to run the compiled Stan executable for the model, link it with the input data and generate posterior samples of parameters, and perform model diagnostics; rstan-stanmodel for a class specifically referred to the compiled model.

Examples

## Not run: 
### A population PK model
m1 <- PMXStanModel(path = tempfile("pk_m1"), compile = TRUE)
print(m1)


## End(Not run)

stanette documentation built on May 11, 2022, 5:11 p.m.