Identifiability

New point release.

What's new?

- Use [PySCeS](https://pysces.github.io) for parameter identifiability analysis of kinetic models using the `CVODE` solver. When performing identifiability analysis in parallel using `multiprocessing`, additional dependencies are required; these can be installed with `pip install "identifiability[pyscesmp]"`.
- Make the degree of the spline used for the profile likelihood plot configurable (default is 2).
- The `params` dictionary is added to the trace dictionary for every step to track all parameters.
- The complete optimization result object is now available from the trace dictionary - useful for troubleshooting individual parameter runs.
- Standard error estimates are no longer required in the input optimization result, as they are not needed for this method.

© Johann M. Rohwer, October 2023

Minor bugfix release.

What's new?

- Fix bugs in multiplot plotting code for CIs.

© Johann M. Rohwer, April 2022

Minor release with some updates.

What's new?

- Implement recalculation of confidence intervals with a different probability, without re-calculating the whole profile likelihood (which is computationally expensive).
- Add Github continuous integration for automatic wheel builds and PyPI upload.

© Johann M. Rohwer, April 2022

What's new?

- Optimization can now be performed in parallel using the `multiprocessing` module.
- Added a method `ConfidenceInterval.plot_all_ci()` to plot confidence intervals for all the parameters analysed.
- First release on PyPI, the module can now be installed simply with `pip install identifiability`.

© Johann M. Rohwer, April 2022

This module performs parameter identifiability analysis to calculate and plot confidence intervals based on a profile-likelihood. The code is adapted from [LMFIT](https://lmfit.github.io/lmfit-py/), with custom functions to select the range for parameter scanning and for plotting the profile likelihood. The significance is assessed with the chi-squared distribution.

What's new?

- add support for using the LMFIT [Model](https://lmfit.github.io/lmfit-py/model.html) class
- make handling of limits more consistent between linear and log parameter scans
- various bug fixes

© Johann M. Rohwer, February 2022

First public release.

This module performs parameter identifiability analysis to calculate and plot confidence intervals based on a profile-likelihood. The code is adapted from [LMFIT](https://lmfit.github.io/lmfit-py/), with custom functions to select the range for parameter scanning and for plotting the profile likelihood. The significance is assessed with the chi-squared distribution.

© Johann M. Rohwer, December 2021