Fits a logistic regression/RSF using glmnet
Function with similar name
Usage
net_logit(
f,
data,
alpha = 1,
penalty.factor = NULL,
type.measure = "deviance",
standardize = TRUE,
na.action = "na.pass",
func = c("glmnet", "cv.glmnet")[1],
...
)
net_rsf(
f,
data,
alpha = 1,
penalty.factor = NULL,
type.measure = "deviance",
standardize = TRUE,
na.action = "na.pass",
func = c("glmnet", "cv.glmnet")[1],
...
)
Arguments
- f
[formula]
Formula of the model to be fitted, with all possible candidate terms.- data
[data.frame,tibble]
Complete data set to be analyzed.#' @param alpha Default is L1-regularization (Lasso regression), withalpha = 1
. L2-regularization (Ridge regression) is done withalpha = 0
, and elastic-net regression is performed for anyalpha
value between0
and1
. For more details, see theglmnet::glmnet()
documentation. For Adaptive and Decay Adaptive Lasso, keepalpha = 1
.- penalty.factor
[numeric,vector=NULL]
Vector of penalty factors to be used for Adaptive Lasso fitting. The vector might have the same length as the the number of columns given by the model matrix,model.matrix(f, data)
. Default isNULL
, in case the same penalty is applied to all variables.- type.measure
[character(1)="deviance"]
Type of measure to evaluate the model internally inglmnet::glmnet()
. For logistic and conditional logistic regression, it is by default"deviance"
.- standardize
[logical(1)=TRUE]
Whether the call toglmnet
should include internal standardization of variables or not. Default is TRUE. It should be set toFALSE
if the variables are already standardized.- na.action
[character(1)="na.pass"]
Default is"na.pass"
, i.e. rows with NAs are not automatically removed from themodel.matrix
used for fitting.- func
[character(1)="glmnet"]{"glmnet", "cv.glmnet"}
The function to be used for fitting. Default isglmnet::glmnet()
. The second option isglmnet::cv.glmnet()
which already performs the cross-validation and might include the variable selection/callibration within.Check option parallel = TRUE from glmnet.