Fast inference over mean and covariance parameters for Generalised Linear Mixed Models
Fast inference over mean and covariance parameters for Generalised Linear Mixed Models.
It implements the mathematical tricks of FaST-LMM for the special case of Linear Mixed Models with a linear covariance matrix and provides an interface to perform inference over millions of covariates in seconds. The Generalised Linear Mixed Model inference is implemented via Expectation Propagation and also makes use of several mathematical tricks to handle large data sets with thousands of samples and millions of covariates.
We recommend installing it via conda:
conda install -c conda-forge glimix-core
Alternatively, glimix-core can also be installed using pip:
pip install glimix-core
The second installation option requires from the user to install liknorm beforehand.
Running the tests
After installation, you can test it
python -c "import glimix_core; glimix_core.test()"
as long as you have pytest.
Here it is a very simple example to get you started:
>>> from numpy import array, ones >>> from numpy_sugar.linalg import economic_qs_linear >>> from glimix_core.lmm import LMM >>> >>> X = array([[1, 2], [3, -1], [1.1, 0.5], [0.5, -0.4]], float) >>> QS = economic_qs_linear(X) >>> X = ones((4, 1)) >>> y = array([-1, 2, 0.3, 0.5]) >>> lmm = LMM(y, X, QS) >>> lmm.fit(verbose=False) >>> lmm.lml() -2.2726234086180557
An extensive documentation of the library can be found at http://glimix-core.readthedocs.org/.
This project is licensed under the MIT License.