gden173 / vspglm

All MATLAB Code and examples related to a Semi-Parametric Vector Generalized Linear Model Solver. The univariate case of this method was first proposed by Huang (2014) in Joint Estimation of the Mean and Error Distribution in Generalized Linear Models https://doi.org/10.1080/01621459.2013.824892

vspglm

• vspglm
  • Description
  • Usage
  • Examples
  • Docs
  • Author

Description

All MATLAB Code and examples related to a Semi-Parametric Vector Generalized Linear Model Solver. The univariate case of this method was first proposed by Huang (2014) in Joint Estimation of the Mean and Error Distribution in Generalized Linear Models https://doi.org/10.1080/01621459.2013.824892

Example of multivariate PMFs for the sorbinil dataset (See Examples)

More information on this function can be found here

Usage

This function uses MATLABs Optimization toolbox and MATLAB >= 2019b.

To use the function, clone the repository and add the function to your MATLABPATH

$ git clone https://github.com/gden173/vspglm.git

Then open this as your current working directory in MATLAB and run the following command in the command window

>> addpath('src', '-end')

This should add the vspglm function to your MATLABPATH variable. This can be checked by running

>> matlabpath

		MATLABPATH
		.
		.
		.

	/path/to/vspglm

Where the path to the cloned directory should be at the end of the matlabpath environment variable.
This is only temporary, and will have to be rerun every MATLAB session. To run the scripts in the examples directory, while in the same directory run

>> addpath('examples', '-end')

from the command window.

Examples

Examples usages can be found in `examples`. The basic usage of the `fit_vspglm` function can be found by searching some of the documentation.

>> help fit_vspglm
  vspglm_mmodel = fit_vspglm(formula, tbl,  links)
  fits a vector generalized semi-parametric linear model and stores
  the model output in vspglmmodel
  the function currently takes 3 arguments,
  formula, a string array of string formulas 
  In the formula argument:
             (y1, y2) ~ (x1, x3) -> y1 and y2 share all regression
             coefficients
             (y1, y2) ~ (x1, (x2&0), (0&x3)) 
             -> y1 ~ x1 + x2, y2 ~ x1 + x3
  Responses can also  have a different number of covariates and still 
  share regression coefficients. 
           (y1, y2, y3) ~ (x1,(x2&x2&0), (0&0&x3))
  This means that 
                 y1 ~ 1  + x1  + x2
                 y2 ~ 1  + x1  + x2
                 y3 ~ 1  + x1  + x3    
 
  Where each variable y1, .., yk, x1, .., xp are columns in 
  the table argument tbl. 
  The last argument is then links, a 1 x k cell array of the link
  functions to be used for each model.

>> 

once the src directory has been added to path.

Docs

For the technical implementation of the model, see the documentation.

Author

• Gabriel Dennis (@gden173)