lasso.cv {lol} | R Documentation |
Cross validation lasso. This function optimizes the lasso solution for correlated regulators by an algorithm. this algorithm chooses the minimum lambda since the penalized package by default use 0 for the minimum, which sometimes take a long time to compute
lasso.cv(y, x=NULL, lambda1=NULL, model='linear', steps=15, minsteps=5, log=TRUE, track=FALSE, standardize= FALSE, unpenalized=~0, nFold=10, nMaxiter = Inf, ...)
y |
A vector of gene expression of a probe, or a list object if x is NULL. In the latter case y should a list of two components y and x, y is a vector of expression and x is a matrix containing copy number variables |
x |
Either a matrix containing CN variables or NULL |
lambda1 |
minimum lambda to use |
model |
which model to use, one of "cox", "logistic", "linear", or "poisson". Default to 'linear' |
steps |
parameter to be passed to penalized |
minsteps |
parameter to be passed to penalized |
log |
parameter to be passed to penalized |
track |
parameter to be passed to penalized |
standardize |
parameter to be passed to penalized |
unpenalized |
parameter to be passed to penalized |
nFold |
parameter to be passed to penalized |
nMaxiter |
parameter to be passed to penalized |
... |
other parameter to be passed to penalized |
A list object of class 'lol', consisting of:
fit |
The final sparse regression fit |
beta |
the coefficients, non-zero ones are significant |
lambda |
the penalty parameter lambda used |
residuals |
regression residuals |
conv |
logical value indicating whether the optimization has converged |
Yinyin Yuan
Goeman, J. J. (2009), L1 penalized estimation in the cox proportional hazards model, Biometrical Journal.
lasso
data(chin07) data <- list(y=chin07$ge[1,], x=t(chin07$cn), nFold=5) res <- lasso.cv(data) res