An R package to fit mixture of linear regressions.
![An example] (./mx1.png)
This package implements and improves an EM algorithm, which can obtain the MLE estimators when the goal is to fit two or more linear regressions through data.
Note that the word "linear" sounds restricting but when one feed in nonlinear transformations of predictors into it, one can fit nonlinear models as well. This is not a big news for experienced users of linear regression.
To install, use the devtools package.
install.packages("devtools")
library(devtools)
devtools::install_github("txzhou/mixtureReg")A [guide] (./Guide.pdf) to use the mixtureReg package is provided.
The already available function regmixEM in the mixtools package can complete a similar job but does not offer the option to impose restrictions to the coefficients.
This causes trouble for researchers who need more flexibility in modeling.
When the situation arises, it might be possible to do some clever data transformation so that to alter the model in order to use the current tools. However, there is a way to solve it more intuitively.
In its lm class, R has already offered powerful coefficient restriction capability through the unique formula representing language.
So this package implements the algorithm based on the powerfulness of the lm class and now offers the same flexibility to model the mixture of regressions.
de Veaux RD (1989). "Mixtures of Linear Regressions." Computational Statistics and Data Analysis, 8, 227-245.
Tatiana Benaglia, Didier Chauveau, David R. Hunter, Derek Young (2009). mixtools: An R Package for Analyzing Finite Mixture Models. Journal of Statistical Software, 32(6), 1-29. URL http://www.jstatsoft.org/v32/i06/.