R/plot-cv-ncvreg.R
plot.cv.ncvreg.Rd
Plots the cross-validation curve from a cv.ncvreg
or
cv.ncvsurv
object, along with standard error bars.
A cv.ncvreg
or cv.ncvsurv
object.
Should horizontal axis be on the log scale? Default is TRUE.
What to plot on the vertical axis. cve
plots the
cross-validation error (deviance); rsq
plots an estimate of the
fraction of the deviance explained by the model (R-squared); snr
plots an estimate of the signal-to-noise ratio; scale
plots, for
family="gaussian"
, an estimate of the scale parameter (standard
deviation); pred
plots, for family="binomial"
, the estimated
prediction error; all
produces all of the above.
If TRUE
(the default), places an axis on top of the
plot denoting the number of variables in the model (i.e., that have a
nonzero regression coefficient) at that value of lambda
.
If TRUE
(the default), draws a vertical line at
the value where cross-validaton error is minimized.
Controls the color of the dots (CV estimates).
Other graphical parameters to plot
Error bars representing approximate 68\
along with the estimates at value of lambda
. For rsq
and
snr
applied to models other than linear regression, the Cox-Snell
R-squared is used.
Breheny P and Huang J. (2011) Coordinate descentalgorithms for nonconvex penalized regression, with applications to biological feature selection. Annals of Applied Statistics, 5: 232-253. c("\Sexpr[results=rd]tools:::Rd_expr_doi(\"#1\")", "10.1214/10-AOAS388")doi:10.1214/10-AOAS388
# Linear regression --------------------------------------------------
data(Prostate)
cvfit <- cv.ncvreg(Prostate$X, Prostate$y)
plot(cvfit)
op <- par(mfrow=c(2,2))
plot(cvfit, type="all")
par(op)
# Logistic regression ------------------------------------------------
data(Heart)
cvfit <- cv.ncvreg(Heart$X, Heart$y, family="binomial")
plot(cvfit)
op <- par(mfrow=c(2,2))
plot(cvfit, type="all")
par(op)
# Cox regression -----------------------------------------------------
data(Lung)
cvfit <- cv.ncvsurv(Lung$X, Lung$y)
op <- par(mfrow=c(1,2))
plot(cvfit)
plot(cvfit, type="rsq")
par(op)