plot.Roc: ROC curves for risk prediction models
In ModelGood: Validation of risk prediction models

Description Usage Arguments Details Value Author(s) See Also Examples

ROC curves for risk prediction models

## S3 method for class 'Roc'
plot(x, ylab = "Sensitivity", xlab = "1-Specificity", models,
  type = "l", shadow = FALSE, simu = FALSE, control, grid = FALSE,
  diag = TRUE, box = FALSE, lwd = 2, lty, col, add = FALSE,
  axes = TRUE, legend, auc, percent = TRUE, ...)

`x`	object obtained with `Roc`
`ylab`	Label y-axis
`xlab`	Label x-axis
`models`	Selection of models to plot. Should be a subset of `names(x$models)`. Makes sense when `x` contains multiple ROC curves.
`type`	The line type
`shadow`	Experimental. Show results of cross-validation.
`simu`	Experimental. Show noinformation results.
`control`	Control which estimates of the ROC curves to draw.
`grid`	If `TRUE` add a grid in the background of the graph.
`diag`	If `TRUE` add a diagonal line.
`box`	If `TRUE` add a box around the graph.
`lwd`	Vector of line widths for the ROC curves.
`lty`	Vector of line types for the ROC curves.
`col`	Vector of colours for the ROC curves.
`add`	If `TRUE` add ROC curves to existing plot.
`axes`	If `TRUE` draw axes.
`legend`	If `TRUE` draw a legend.
`auc`	If `TRUE` add the area under the curve to the legend.
`percent`	If `TRUE` show percent axes.
`...`	Use for smart control of some plot elements.

Multiple ROC curves are shown in one graph.

ROC curves

Thomas A. Gerds <tag@biostat.ku.dk>

Roc

# generate som data
set.seed(40)
N=40
Y=rbinom(N,1,.5)
X1=rnorm(N)
X1[Y==1]=rnorm(sum(Y==1),mean=rbinom(sum(Y==1),1,.5))
X2=rnorm(N)
X2[Y==0]=rnorm(sum(Y==0),mean=rbinom(sum(Y==0),1,.5))
dat <- data.frame(Y=Y,X1=X1,X2=X2)

# fit two logistic regression models
lm1 <- glm(Y~X1,data=dat,family="binomial")
lm2 <- glm(Y~X2+X1,data=dat,family="binomial")
plot(Roc(list(lm1,lm2),data=dat))

# add the area under the curves

plot(Roc(list(lm1,lm2),data=dat),auc=TRUE)

# alternatively, one can directly work with formula objects:
plot(Roc(list(LR.X1=Y~X1,LR.X1.X2=Y~X2+X1),data=dat),auc=TRUE)

# beyond the logistic regression model.
# the following example is optimized for speed
# illustrating the syntax,
# and not for optimized for performance of the
# randomForest or elastic net
library(randomForest)
library(glmnet)
dat$Y=factor(dat$Y)
rf <- randomForest(Y~X1+X2,data=dat,ntree=10)
en <- ElasticNet(Y~X1+X2,data=dat,nfolds=10,alpha=0.1)
set.seed(6)
rocCV=Roc(list(RandomForest=rf,ElasticNet=en,LogisticRegression=lm2),
  data=dat,
  verbose=FALSE,
  splitMethod="bootcv",
  B=4,
  cbRatio=1)
plot(rocCV,yaxis.las=2,legend.title="4 bootstrap-crossvalidation steps")

randomForest 4.6-14
Type rfNews() to see new features/changes/bug fixes.
Loading required package: Matrix
Loading required package: foreach
Loaded glmnet 2.0-16