#	Name	Function	Parameter	R package
1	all subsets regression, maximizing the Bayesian information criterion (BIC)	regsubsets	nvmax=15, nbest=1, method=“forward”, really.big=T	leaps
2	stepwise regression, maximizing the BIC	stepAIC	−	MASS
3	stepwise regression, maximizing the Akaike information criterion (AIC)	stepAIC	−	MASS
4	Lasso regression	cv.glmnet	family=“gaussian”, nfolds=10, alpha = 1	glmnet
5	multivariate adaptive regression splines (MARS)	earth	degree = 1, trace = 0, nk = 500	earth
6	Random Forest	randomForest	−	randomForest
7	principal component regression (PCR)	pcr	ncomp = 5 (during prediction)	pls
8	Partial Least Squares (PLSR)	plsr	ncomp = 5 (during prediction)	pls
9	Support Vector Machine w/ L1 loss function (SMV+L1)	tune	method=svm, ranges = list(epsilon=seq(0,1,0.025), cost=2^(2:8)), kernel=“radial”	e1071