pharmpy.modeling.evaluate_eta_gradient#

pharmpy.modeling.evaluate_eta_gradient(model, etas=None, parameters=None, dataset=None)[source]#

Evaluate the numeric eta gradient

The gradient is evaluated at the current model parameter values or optionally at the given parameter values. The gradient is done for each data record in the model dataset or optionally using the dataset argument. The gradient is done at the current eta values or optionally at the given eta values.

This function currently only support models without ODE systems

Parameters:

model (Model) – Pharmpy model
etas (dict) – Optional dictionary of eta values
parameters (dict) – Optional dictionary of parameters and values
dataset (pd.DataFrame) – Optional dataset

Returns:

pd.DataFrame – Gradient

Examples

>>> from pharmpy.modeling import load_example_model, evaluate_eta_gradient
>>> from pharmpy.tools import load_example_modelfit_results
>>> model = load_example_model("pheno_linear")
>>> results = load_example_modelfit_results("pheno_linear")
>>> etas = results.individual_estimates
>>> evaluate_eta_gradient(model, etas=etas)
     dF/dETA_1  dF/dETA_2
1     -0.159537 -17.609116
2     -9.325893 -19.562289
3     -0.104417 -11.346161
4     -4.452951 -16.682310
5    -10.838840 -18.981836
..          ...        ...
151   -5.424423 -19.973013
152  -14.497185 -17.344797
153   -0.198714 -22.697161
154   -7.987731 -23.941806
155  -15.817067 -22.309945

[155 rows x 2 columns]