cv.step.plr {stepPlr} | R Documentation |
This function computes cross-validated deviance or prediction errors
for step.plr.
The parameters that can be cross-validated are
lambda
and cp
.
cv.step.plr(x, y, weights = rep(1, length(y)), nfold = 5, folds = NULL, lambda = c(1e-4, 1e-2, 1), cp = c("aic", "bic"), cv.type=c("deviance", "class"), trace = TRUE, ...)
x |
matrix of features |
y |
binary response |
weights |
an optional vector of weights for observations |
nfold |
number of folds to be used in cross-validation. Default is
nfold=5.
|
folds |
the list of cross-validation folds. Its length must be nfold. If
NULL, the folds are randomly generated.
|
lambda |
vector of the candidate values for lambda in step.plr
|
cp |
vector of the candidate values for cp in step.plr
|
cv.type |
If cv.type=deviance, cross-validated deviances are returned. If
cv.type=class, cross-validated prediction errors are returned.
|
trace |
If TRUE, the steps are printed out.
|
... |
other options for step.plr
|
This function computes cross-validated deviance or prediction errors
for step.plr.
The parameters that can be cross-validated are
lambda
and cp
. If both are input as vectors (of length
greater than 1), then a two-dimensional cross-validation is done. If
either one is input as a single value, then the cross-validation is
done only on the parameter with multiple inputs.
Mee Young Park and Trevor Hastie
Mee Young Park and Trevor Hastie (2006) Penalized Logistic Regression for Detecting Gene Interactions - available at the authors' websites, http://stat.stanford.edu/~mypark or http://stat.stanford.edu/~hastie/pub.htm.
step.plr
n <- 100 p <- 5 x <- matrix(sample(seq(3),n*p,replace=TRUE),nrow=n) y <- sample(c(0,1),n,replace=TRUE) level <- vector("list",length=p) for (i in 1:p) level[[i]] <- seq(3) cvfit1 <- cv.step.plr(x,y,level=level,lambda=c(1e-4,1e-2,1),cp="bic") cvfit2 <- cv.step.plr(x,y,level=level,lambda=1e-4,cp=c(2,3,4)) cvfit3 <- cv.step.plr(x,y,level=level,lambda=c(1e-4,1e-2,1),cp=c(2,3,4))