estimagic is a Python package for nonlinear optimization with or without constraints. It is particularly suited to solve difficult nonlinear estimation problems. On top, it provides functionality to perform statistical inference on estimated parameters.
- estimagic wraps algorithms from scipy.optimize, nlopt, pygmo and more.
- estimagic implements constraints efficiently via reparametrization, so you can solve constrained problems with any optimizer that supports bounds.
- The parameters of an optimization problem can be arbitrary pytrees
- The complete history of parameters and function evaluations can be saved in a database for maximum reproducibility.
- Painless and efficient multistart optimization.
- The progress of the optimization is displayed in real time via an interactive dashboard.
- You can estimate a model using method of simulated moments (MSM), calculate standard errors and do sensitivity analysis with just one function call.
- Asymptotic standard errors for maximum likelihood estimation.
- estimagic also provides bootstrap confidence intervals and standard errors. Of course the bootstrap procedures are parallelized.
- estimagic can calculate precise numerical derivatives using Richardson extrapolations.
- Function evaluations needed for numerical derivatives can be done in parallel with pre-implemented or user provided batch evaluators.
The package can be installed via conda. To do so, type the following commands in a terminal:
$ conda config --add channels conda-forge
$ conda install estimagicThe first line adds conda-forge to your conda channels. This is necessary for conda to find all dependencies of estimagic. The second line installs estimagic and its dependencies.
Only scipy is a mandatory dependency of estimagic. Other algorithms become available
if you install more packages. We make this optional because most of the time you will
use at least one additional package, but only very rarely will you need all of them.
For an overview of all optimizers and the packages you need to install to enable them
see {ref}list_of_algorithms.
To enable all algorithms at once, do the following:
conda install nlopt
pip install Py-BOBYQA
pip install DFO-LS
conda install petsc4py (Not available on Windows)
conda install cyipopt
conda install pygmo
pip install fides>=0.7.4 (Make sure you have at least 0.7.1)
The documentation is hosted (on rtd)
If you use Estimagic for your research, please do not forget to cite it.
@Unpublished{Gabler2022,
Title = {A Python Tool for the Estimation of large scale scientific models.},
Author = {Janos Gabler},
Year = {2022},
Url = {https://github.com/OpenSourceEconomics/estimagic}
}