EstimationStep#

class pharmpy.model.EstimationStep(method, interaction=False, parameter_uncertainty_method=None, evaluation=False, maximum_evaluations=None, laplace=False, isample=None, niter=None, auto=None, keep_every_nth_iter=None, residuals=None, predictions=None, solver=None, solver_rtol=None, solver_atol=None, tool_options=None, eta_derivatives=None, epsilon_derivatives=None)[source]#

Bases: Step

Definition of one estimation operation

Attributes Summary

`auto`	Let estimation tool automatically add settings
`epsilon_derivatives`	List of names of epsilons for which to calculate derivatives
`eta_derivatives`	List of names of etas for which to calculate derivatives
`evaluation`	Only perform model evaluation
`interaction`	Preserve eta-epsilon interaction in the computation of the objective function
`isample`	Number of samples per subject (or similar) for EM methods
`keep_every_nth_iter`	Keep results for every nth iteration
`laplace`	Use the laplacian method
`maximum_evaluations`	Maximum allowable number of evaluations of the objective function
`method`	Name of the estimation method
`niter`	Number of iterations for EM methods
`parameter_uncertainty_method`	Method to use when estimating parameter uncertainty.
`predictions`	List of predictions to estimate
`residuals`	List of residuals to calculate
`supported_methods`
`supported_parameter_uncertainty_methods`
`tool_options`	Dictionary of tool specific options

Methods Summary

`create`(method[, interaction, ...])
`from_dict`(d)
`replace`(**kwargs)	Derive a new EstimationStep with new properties
`to_dict`()

Attributes Documentation

auto#: Let estimation tool automatically add settings

epsilon_derivatives#: List of names of epsilons for which to calculate derivatives

eta_derivatives#: List of names of etas for which to calculate derivatives

evaluation#: Only perform model evaluation

interaction#: Preserve eta-epsilon interaction in the computation of the objective function

isample#: Number of samples per subject (or similar) for EM methods

keep_every_nth_iter#: Keep results for every nth iteration

laplace#: Use the laplacian method

maximum_evaluations#: Maximum allowable number of evaluations of the objective function

method#: Name of the estimation method

niter#: Number of iterations for EM methods

parameter_uncertainty_method#

Method to use when estimating parameter uncertainty. Supported methods and their corresponding NMTRAN code:

Method	NMTRAN
Sandwich	$COVARIANCE
Cross-product gradient (SMAT)	$COVARIANCE MATRIX=S
Observed FIM (RMAT)	$COVARIANCE MATRIX=R
Expected FIM (EFIM)	$DESIGN

By default the following options are appended: UNCONDITIONAL: The uncertainty step is implemented regardless of how the estimation step terminates. PRINT=E: Print the eigenvalues of the correlation matrix. PRECOND=1: Perform up to 1 preconditioning cycle on the R matrix.

predictions#: List of predictions to estimate

residuals#: List of residuals to calculate

supported_methods = frozenset({'BAYES', 'FO', 'FOCE', 'IMP', 'IMPMAP', 'ITS', 'SAEM'})#

supported_parameter_uncertainty_methods = frozenset({'EFIM', 'RMAT', 'SANDWICH', 'SMAT'})#

tool_options#: Dictionary of tool specific options

Methods Documentation

classmethod create(method, interaction=False, parameter_uncertainty_method=None, evaluation=False, maximum_evaluations=None, laplace=False, isample=None, niter=None, auto=None, keep_every_nth_iter=None, residuals=None, predictions=None, solver=None, solver_rtol=None, solver_atol=None, tool_options=None, eta_derivatives=None, epsilon_derivatives=None)[source]#

classmethod from_dict(d)[source]#

replace(**kwargs)[source]#: Derive a new EstimationStep with new properties

to_dict()[source]#