This package allows to calculate the cross-validated log model evidence (cvLME) [1,2,3] in multiple programming languages.
Based on calculated cvLMEs, cross-validated Bayesian model selection (cvBMS) within a model space can be performed.
Currently, it supports the following model structures
MS= model space; for general model selection operations;GLM= univariate general linear model; for linear regression;Poiss= Poisson distribution with exposures; for count data;
which are implemented in the following languages
LaTeX: for documentation purposes only, written in TeXstudio 2.8.8;MATLAB: developed in and compatible with MATLAB R2013b;Python: developed in and compatible with Python 3.7.
Extensive documentation is given in the manual accompanying this repository [4].
In the following examples, <name-of-the-model-class> is either "GLM" or "Poiss".
To use the module, it is simply imported via import cvBMS, e.g. at the beginning of your analysis script.
In a Python console, type help(cvBMS) and help(cvBMS.<name-of-the-model-class>) to learn more.
Please also read the implementation notes in LaTeX\cvBMS.pdf [4] to apply in Python.
To use these functions, simply rename and put the sub-directory MATLAB into your MATLAB path.
In the command window, type help <name-of-the-model-class>_cvLME.m to learn more.
Please also read the implementation notes in LaTeX\cvBMS.pdf [4] to apply in MATLAB.
Simply open <name-of-the-model-class>.tex in sub-directory LaTeX to access and reuse formulas.
Please open LaTeX\cvBMS.pdf [4] to view the PDF output from this LaTeX code.
[1] https://www.sciencedirect.com/science/article/pii/S1053811916303615
[2] https://www.sciencedirect.com/science/article/pii/S105381191730527X
[3] https://www.sciencedirect.com/science/article/pii/S0165027018301468
[4] https://github.com/JoramSoch/cvLME/blob/master/LaTeX/cvBMS.pdf