Prev 301343 / 398503 Next

optimal cut off with Epi package

Frank E Harrell Jr

Thu, Jul 26, 2012 6:37 AM

It is not appropriate to seek a cutoff for a test unless every patient is
known to have exactly the same utility function, which is extremely rare. 
There is nothing wrong with developing probability models and stopping at
that point, deferring the decision to the possessor of the utility function.
Frank

Luigi wrote

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 an 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=0.431. would
be
this the optimal cut off? Otherwise I 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]]

______________________________________________
R-help@ mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.

-----
Frank Harrell
Department of Biostatistics, Vanderbilt University
--
View this message in context: http://r.789695.n4.nabble.com/optimal-cut-off-with-Epi-package-tp4637904p4637932.html
Sent from the R help mailing list archive at Nabble.com.

Thread (2 messages)

Luigi Marongiu optimal cut off with Epi package Jul 26 Frank E Harrell Jr optimal cut off with Epi package Jul 26