ROC – plot

November 1, 2010

(This article was first published on Ecological Modelling... » R, and kindly contributed to R-bloggers)

There are many implantation in R already of ROC plots (e.g. in the packages PresenceAbsence, ROCR). I just wrote my own very simple script just to get a better understanding of it.

## ROC - plot
d <- data.frame(id=1:100, ob=sample(c(1,0), 100, replace=T), m1=sample(seq(0,1,by=0.01), 100, replace=T))
# interval to calculate the threshold
int <- 100
th <- seq(0,1, length=int)
roc.plot <- data.frame(sen=rep(NA,int), spe=rep(NA,int))

for (i in 1:int)
 # get tn, tp, fn, fp
 tn <- nrow(d[d[,3]<th[i]&d[,2]==0,])
 fn <- nrow(d[d[,3]<th[i]&d[,2]==1,])
 fp <- nrow(d[d[,3]>th[i]&d[,2]==0,])
 tp <- nrow(d[d[,3]>th[i]&d[,2]==1,])

 # sensitivity, if sensitivty == 1, everything all positives are found
 roc.plot[i,'sen'] <- tp/(tp+fn)
 # specificity, if specificity == 1, all negatives are found
 roc.plot[i,'spe'] <- tn/(tn+fp)
with(roc.plot, plot(1-spe, sen, type="l"))

To leave a comment for the author, please follow the link and comment on their blog: Ecological Modelling... » R. offers daily e-mail updates about R news and tutorials on topics such as: visualization (ggplot2, Boxplots, maps, animation), programming (RStudio, Sweave, LaTeX, SQL, Eclipse, git, hadoop, Web Scraping) statistics (regression, PCA, time series, trading) and more...

If you got this far, why not subscribe for updates from the site? Choose your flavor: e-mail, twitter, RSS, or facebook...

Comments are closed.