elnet | R Documentation |
Compute least squares EN estimates for linear regression with optional observation weights and penalty loadings.
elnet(
x,
y,
alpha,
nlambda = 100,
lambda_min_ratio,
lambda,
penalty_loadings,
weights,
intercept = TRUE,
en_algorithm_opts,
sparse = FALSE,
eps = 1e-06,
standardize = TRUE,
correction = deprecated(),
xtest = deprecated(),
options = deprecated()
)
x |
|
y |
vector of response values of length |
alpha |
elastic net penalty mixing parameter with |
nlambda |
number of penalization levels. |
lambda_min_ratio |
Smallest value of the penalization level as a fraction of the largest
level (i.e., the smallest value for which all coefficients are zero).
The default depends on the sample size relative to the number of variables and |
lambda |
optional user-supplied sequence of penalization levels.
If given and not |
penalty_loadings |
a vector of positive penalty loadings (a.k.a. weights) for different penalization of each coefficient. |
weights |
a vector of positive observation weights. |
intercept |
include an intercept in the model. |
en_algorithm_opts |
options for the EN algorithm. See en_algorithm_options for details. |
sparse |
use sparse coefficient vectors. |
eps |
numerical tolerance. |
standardize |
standardize variables to have unit variance. Coefficients are always returned in original scale. |
correction |
defunct. Correction for EN estimates is not supported anymore. |
xtest |
defunct. |
options |
deprecated. Use |
The elastic net estimator for the linear regression model solves the optimization problem
argmin_{\mu, \beta}
(1/2n) \sum_i w_i (y_i - \mu - x_i' \beta)^2 +
\lambda \sum_j 0.5 (1 - \alpha) \beta_j^2 + \alpha l_j |\beta_j|
with observation weights w_i
and penalty loadings l_j
.
a list-like object with the following items
alpha
the sequence of alpha
parameters.
lambda
a list of sequences of penalization levels, one per alpha
parameter.
estimates
a list of estimates. Each estimate contains the following information:
intercept
intercept estimate.
beta
beta (slope) estimate.
lambda
penalization level at which the estimate is computed.
alpha
alpha hyper-parameter at which the estimate is computed.
statuscode
if > 0
the algorithm experienced issues when
computing the estimate.
status
optional status message from the algorithm.
call
the original call.
pense()
for an S-estimate of regression with elastic net penalty.
coef.pense_fit()
for extracting coefficient estimates.
plot.pense_fit()
for plotting the regularization path.
Other functions for computing non-robust estimates:
elnet_cv()
# Compute the LS-EN regularization path for Freeny's revenue data
# (see ?freeny)
data(freeny)
x <- as.matrix(freeny[ , 2:5])
regpath <- elnet(x, freeny$y, alpha = c(0.5, 0.75))
plot(regpath)
plot(regpath, alpha = 0.75)
# Extract the coefficients at a certain penalization level
coef(regpath, lambda = regpath$lambda[[1]][[5]],
alpha = 0.75)
# What penalization level leads to good prediction performance?
set.seed(123)
cv_results <- elnet_cv(x, freeny$y, alpha = c(0.5, 0.75),
cv_repl = 10, cv_k = 4,
cv_measure = "tau")
plot(cv_results, se_mult = 1.5)
plot(cv_results, se_mult = 1.5, what = "coef.path")
# Extract the coefficients at the penalization level with
# smallest prediction error ...
summary(cv_results)
coef(cv_results)
# ... or at the penalization level with prediction error
# statistically indistinguishable from the minimum.
summary(cv_results, lambda = "1.5-se")
coef(cv_results, lambda = "1.5-se")
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.