[R] Re : ROC Analysis

Hi Camille, Probably you have to check wether there is any infinte value in x. Or calculate something like that for your x-axis: x[1:(ll-1)]+diff(x)/2 Regards, Pascal - Mail original - De : Camille Leclerc À : r-help@r-project.org Cc : Envoyé le : Lundi 26 mars 2012 0h32 Objet : Re:

[R] Re : ROC Analysis

Hi Camille, Does following work? ll <- length(x) (TPR[1:(ll-1)]-TPR[2:ll])/(FPR[1:(ll-1)]-FPR[2:ll]) Regards, Pascal -- View this message in context: http://r.789695.n4.nabble.com/ROC-Analysis-tp4469203p4478233.html Sent from the R help mailing list archive at Nabble.com. ___

[R] Re : ROC Analysis

Hi Camille, Probably by adding these lines: fpf <- unlist(performance(pred,"fpr")@y.values) fnf <- unlist(performance(pred,"fnr")@y.values) pLhood <- tpf/fpf   # Positive Likelihood Ratio nLhood <- fnf/tnf   # Negative Likelihood Ratio par(mfrow=c(1,2)) plot(x, pLhood, t='l', xlab='Value', ylab

[R] Re : ROC Analysis

Hi Camille, I am surprised by your answer. If you do: pred <- prediction(ROCR.simple$predictions*1000, ROCR.simple$labels) x <- rev(unlist(pred@cutoffs)) You can see than x values are now between 0 and 1000. So, it should be probably the same for your data. Regards, Pascal -- View this me

[R] Re : ROC Analysis

Hi Camille, Do you need something like that? ### library(ROCR) data(ROCR.simple) pred <- prediction(ROCR.simple$predictions, ROCR.simple$labels) tpf <- unlist(performance(pred,"tpr")@y.values) tnf <- unlist(performance(pred,"tnr")@y.values) x <- rev(unlist(pred@cutoffs)) plot(x