bestsubset
vignettes/install_bestsubset.Rmd
install_bestsubset.RmdThe APES method relies on a best subset algorithm for linear
regression model. In cases where the number of regressors is less than
30, the leaps
package is sufficiently fast to provide a result in a reasonable
time. For cases where the number of regressors becomes larger, we
recommend using the bestsubset
package as the backend. However, the installation of this package is
not quite straight forward. The purpose of this vignette is to detail
some key steps to install this package.
bestsubset
The bestsubset package that APES depends on
can be installed from GitHub via the R command
devtools::install_github("ryantibs/best-subset/bestsubset").
While this step is relatively straight forward, the real hurdle is the
installation of the gurobi R package which the
commercial company Gurobi has copyright over.
To install gurobi, follow the steps in this
vignette.
In general, you should:
Register an account with Gurobi at https://www.gurobi.com/login/.
Review the licensing options at http://www.gurobi.com/products/licensing-pricing/licensing-overview. Academic affiliates can obtain a free license from https://www.gurobi.com/downloads/end-user-license-agreement-academic/.
Open up your system terminal and enter
grbgetkey <license-id>, where your
<license-id> is in the form of
XXXXXXXX-XXXX-XXXX-XXXX-XXXXXXXXXXXX from step 2.
Download the Gurobi software from https://www.gurobi.com/downloads/ and select the software version and operating system. Documentations specific to your operating system can be found at http://www.gurobi.com/documentation/.
Once the software is downloaded, install it at your desired path.
The R package will be in a subfolder under the installation
path. For example, on MacOS, the default installation path for Gurobi
9.0.1 is /Library/gurobi901 and the R command
for installing gurobi is
install.packages('/Library/gurobi901/mac64/R/gurobi_9.0-1_R_3.6.1.tgz', repos=NULL).
You will also need install the slam package via
install.packages("slam", repos = "https://cloud.r-project.org").
Test if gurobi is installed by running
library(gurobi) in R.
Finally, the bestsubset package can be installed
from GitHub via the R command
devtools::install_github("ryantibs/best-subset/bestsubset").