BayesMSD: properly fitting MSDs

While inspection of MSD curves is one of the most ubiquitous ways of analyzing particle tracking data, it is also well known that extracting model parameters from MSD curves is a statistical minefield¹. This problem can be addressed quite nicely in the language of Gaussian processes, allowing statistically rigorous MSD fits. This provides, for example, error bars on estimated model parameters, which are quite noticeably missing from the current literature.

For a Quickstart intro, more extensive Tutorials & Examples and the full API reference refer to the documentation hosted at ReadTheDocs.

To install bayesmsd you can use the latest stable version from PyPI

$ pip install --upgrade bayesmsd

or the very latest updates right from GitHub:

$ pip install git+https://github.com/OpenTrajectoryAnalysis/bayesmsd

When cloning the repo and installing in editable mode, make sure to use make setup to setup the parts of the local environment that are not tracked in git (see Developers):

$ git clone https://github.com/OpenTrajectoryAnalysis/bayesmsd
$ cd bayesmsd && make setup
$ pip install -e .

Developers

We use GNU make to automate recurrent tasks. Targets include:

• make setup : set up the local environment after cloning. Requires nbstripout which is used to remove output and empty cells from the example notebooks.
• make recompile : (re-)compile cython code
• make build : build wheels for distribution on PyPI
• make tests : run unittests
• make docs : build Sphinx documentation

Footnotes

1. Vestergaard, Blainey, Flyvbjerg, Optimal estimation of diffusion coefficients from single-particle trajectories, Physical Review E, 2014; DOI ↩