Yet another linear regression library.
scattr is a plug-n-play tool for performing linear regression in the presence of intrinsic scatter and for measurements with uncertainties on both the dependent and independent variables.
This is built upon other publicly available tools (specifically, linmix1 and LIRA2), but provides a more statistically consistent approach for dealing with observational uncertainties. More details will be provided asap.
To get scattr running on your computer, it should be enough to run
python -m pip install git+https://github.com/lucadimascolo/scattr.git
This will download and install the latest version of scattr as well as all the required dependencies.
Currently, there is a minor issue with the latest pip-installable version of numpyro, as its nested sampler depends on an older version of jaxns. If you are planning to run a nested sampling, I suggest to install the latest dev-release of numpyro as
python -m pip install git+https://github.com/pyro-ppl/numpyro.git
A preliminary example can be found in init.py.
1 "Some Aspects of Measurement Error in Linear Regression of Astronomical Data", Brandon C. Kelly, ApJ, 665, 1489 (2007)
2 "A Bayesian approach to linear regression in astronomy", Mauro Sereno, MNRAS, 455, 2149 (2016)