guhjy / matrixpls

R package plspm

matrixpls

matrixpls is a matrix based implementation of Partial Least Squares Path Modeling algorithm. The matrix based implementation is computationally more efficient than existing PLS implementation (plspm and semPLS) and does not require raw data but calculates the PLS estimates from a covariance matrix. The package is designed towards Monte Carlo simulations with simsem.

matrixpls is currently under development and not suitable for end-users.

Development

matrixpls has three development targets or main requirements

1. matrixpls should be able to estimate all relevant vignettes from simsem
2. matrixpls works as drop-in replacement for the most recent version of plspm
3. matrixpls works as drop-in replacement for the most recent version of plsSEM

All simsem vignettes should work by just replacing the call to sim with a call to matrixpls.sim. Compatibility with plspm and semPLS should be as easy as substuting the calls to plspm function with matrixpls.plspm and semPLS with matrixpls.semPLS.

matrixpls uses test driven development with Runit. This means that the tests are written first and then software is implemented to conform with the tests.

Requirements and Installation

To install the latest development version of matrixpls from github (using the package "devtools"), run in your R console:

# install.packages("devtools") 
library(devtools)
install_github("mronkko/matrixpls")

To check out the source code and install from local copy, run the following shell commands

git clone https://github.com/mronkko/matrixpls.git
R CMD BUILD matrixpls
R CMD CHECK matrixpls_*.tar.gz
R CMD INSTALL matrixpls_*.tar.gz

You may need to install development versions of Lavaan and SimSem to run the development verison of matrixpls

 install.packages("lavaan", repos="http://www.da.ugent.be", type="source")

 install.packages("simsem", repos="http://rweb.quant.ku.edu/kran", type="source")

Author Contact

Mikko Rönkkö ([email protected])