Learn R Programming

⚠️There's a newer version (1.2.17) of this package.Take me there.

CorReg (version 1.0.5)

Linear Regression Based on Linear Structure Between Covariates

Description

Linear regression based on a recursive structural equation model (explicit correlations) found by a MCMC algorithm. It permits to face highly correlated covariates. Variable selection is included (by lasso, elastic net, etc.). It also provides some graphical tools for basic statistics and regression trees.

Copy Link

Version

Install

install.packages('CorReg')

Monthly Downloads

167

Version

1.0.5

License

CeCILL

Maintainer

Clement THERY

Last Published

June 9th, 2015

Functions in CorReg (1.0.5)

mixture_generator

Gaussian mixtures dataset generator with regression between the covariates

Matplot with curves comparison by background colors.

Destructing values to have missing ones

read the structure and explain it

Compute the BIC of a given structure

density_estimation

BIC of estimated marginal gaussian mixture densities

Quick tutorial for CorReg package

To show the missing values of a dataset

simple MSE function

structureFinder

MCMC algorithm to find a structure between the covariates

Removes missing values (rows and column to obtain a large full matrix)

To compare sub-regression structures

decision tree in a recursive way

Linear regression using CorReg's method, with variable selection.

plot and give confidence intervals on the coefficients estimated in a model or for proportions

Sparse structure research

Give the partition implied by a structure

Computes the MSE on the joint distribution of the dataset

Selection method based on p-values (coefficients)

To clean Z based on R2

Fill the missing values in the dataset

compare 0 values in two vectors

Upgrades a package to the lastest version on R-forge

a summary-like function

Curve of the BIC for each possible p2 with a fixed Z and truncature of Z

clean Z columns (if BIC improved)

Estimates R2 of each subregression

Imputation of missing values knowing alpha (E step of the EM)

Cross validation

Response variable generator with a linear model

Probability of Z without knowing the dataset. It also gives the exact number of binary nilpotent matrices of size p.

compare signs of the coefficients in two vectors

compare signs of the coefficients in two vectors

Estimates B matrix

clean Z (if BIC improved)

Ordinary Least Square efficiently computed with SEM for missing values

initialization based on a wheight matrix (correlation or other)