Creating an IQRsysProject object

This function takes and IQRsysEst object and other optional input arguments and exports an NMLE project to a folder. It is this folder that we call an "IQRsysProject" object and it is tool specific (e.g. NONMEM, MONOLIX, or NLMIXR). The original IQRmodel used here will be exported to the project folder as well as "model.txt"

IQRsysProject(
  est,
  projectPath,
  keepProjectFolder = FALSE,
  SIMOPT.method = "lsodes",
  SIMOPT.atol = 1e-06,
  SIMOPT.rtol = 1e-06,
  SIMOPT.hmin = 0,
  SIMOPT.hmax = NULL,
  SIMOPT.hini = 0,
  SIMOPT.maxsteps = 5000,
  SIMOPT.nauxtimes = 0,
  SIMOPT.cores = 1,
  opt.method = "trust",
  opt.nfits = 1,
  opt.sd = 1,
  opt.rinit = 1,
  opt.rmax = 10,
  opt.iterlim = 100,
  opt.prior_sigma = 10,
  opt.parlower = NULL,
  opt.parupper = NULL,
  algOpt.SEED = 123456,
  FLAGprofileLL = F,
  FLAGkeepFits = F,
  FLAGchecks = T,
  ...
)

Arguments

est

IQRsysEst object

projectPath

Path where to generate the NLME project

keepProjectFolder

if FALSE (default), remove existing project folder before writing the project again

SIMOPT.method

character denoting the integration method, e.g. "lsodes" (default), "lsoda" See the deSolve package for a list of available integrators

SIMOPT.atol

numeric, absolute error tolerance

SIMOPT.rtol

numeric, relative error tolerance

SIMOPT.hmin

numeric, minimal integration step size

SIMOPT.hmax

numeric, maximal integration step size

SIMOPT.hini

numeric, initial integration step size

SIMOPT.maxsteps

numeric, maximum number of integration steps

SIMOPT.nauxtimes

additional times to simulate the prediction function

SIMOPT.cores

number of cores used for parallel evaluation of conditions/individuals

opt.method

optimization method ("trust" = trust region optimization using sensitivity equations, "hjkb" = derivative-free optimization by the Hooke-Jeeves algorithm, also known as pattern search, or "nmkb" = derivative-free optimization by the Nelder-Mead algorighm).

opt.nfits

number of fits (starting from random positions)

opt.sd

the standard deviation of the initial parameter sample from where optimization is started

opt.rinit

initial trust region radius for optimization

opt.rmax

maximal trust region radius for optimization

opt.iterlim

iteration limit for optimization

opt.prior_sigma

use a quadratic prior to regularize the estimation problem. Larger sigma values correspond to a weaker prior. The prior sigma should be much larger than the expected parameter uncertainty. As a rule of thumb, "1" corresponds to a rather strong prior whereas "10" and larger corresponds to a weak prior. To switch the prior off, use NULL as value.

opt.parlower, opt.parupper

named vector providing lower/upper bounds for parameter values. Provide values on the linear (natural) scale, even if parameters are estimated on log-scale.

algOpt.SEED

set seed for fully reproducible optimization outcome.

FLAGprofileLL

(Default: FALSE) Compute profile likelihood if TRUE AND if best fit is converged. Do not use FLAGprofileLL = TRUE with opt.prior_sigma = NULL unless the model is fully identifiable.

FLAGkeepFits

When post-processing fits, keep original parlist output of mstrust? If FALSE, potential error messages might be deleted.

FLAGchecks

Perform checks of the generated functions before running fits (Default: TRUE). For very large systems, this step can be time-consuming in which case it might be switched off by the user.

...

currently not used

Value

An IQRsysProject object

See also