Dear all,
I would like to calculate the optimal cut off (threshold) of a test using
the Epi package. Here I am presenting some data based on the output of two
tests. I am interested in identifying the optimal cut off and its 95% CI.
Running the ROC() function with the Epi package I obtain a nice picture that
returns what I interpret as the optimal cut off with lr.eta=0.491. would be
this the optimal cut off? Otherwise how can I calculate it? And its
confidence interval?
Thank you,
Luigi Marongiu, MSc
####################################
test1<-c( 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
1, 1, 1, 1, 1, 1, 1)
test2<-c( 0.65662706, 0.009706075, 0.003134876, 0.049831384,
1.089876198, 0.136582023, 0.165812622, 1.316512535, 0.725196908,
6.214688703, 48.26356415, 0.004775112, 0.035568226, 0.36934994,
0.452627677, 1.001904226, 0.116876181, 0.127614619, 1.345812733,
1.487365838, 4.321874201, 0.373389928, 0.002137676, 0.019958462,
43.51470584, 10.07600936, 8.668998078, 11.2505088, 45.53510205,
49.80855616, 1.375550815, 9.798996492, 9.494694175, 23.24171357,
0.779040988, 664.0886249, 15.05400384, 2.365265177, 0.076619211,
0.692909116, 8.497272898, 15.57700003, 20.83909961, 0.833613282,
2.84624862, 0.46118499, 7.330049094, 1.612815795, 3.695709614,
17.75107595)
### create data frame from matrix
my.mat<-matrix(c(test1, test2), nrow=50, byrow=FALSE)
dimnames(my.mat)<-list(c(1:50),c("class", "test"))
my.data<-as.data.frame(my.mat)
attach(my.data)
### LOAD PACKAGE
library(Epi)
### ROC analysis
ROC(form=my.data$class ~ my.data$test, plot="ROC",
data=my.data,
main="ROC with Epi package", MI=TRUE, MX=TRUE, PV=TRUE)
### REPORT: lr.eta=0.491
[[alternative HTML version deleted]]
