[R] Extracting information from ROC??

svannoy steven.vannoy at umb.edu
Tue Apr 29 21:47:30 CEST 2014


I've used the ROC function from the Epi package to plot an ROC curve and
generate descriptive statistics (i.e. sensitivity, specificity, PPV, NPV)
for what ROC considers the best predictor score - this is done automatically
by ROC on the graph. What I want to do, is get these same descriptives for
other predictor scores.

Here is the code I've used:

ROC(stat=diseaseStatus, test=screenScore, cuts=c(3,4,6), PV=T)

This prints out the ROC curve. From prior exploration I know that ROC
identifies screenScore=5 as the "best" predictor score (sens=91.7%,
spec=88.9%, PPV=0.2%, NPV = 82.5%, I have a very low base rate disease).
When I use the "cuts" option, ROC adds those values to the graph (3.00,
4.00, 6.00) but does not print the descriptives for those predictor values.

I don't necessarily want them on the graph, I'm happy to just have the
descriptives for '5' on there, but I want to have access to the descriptives
for other values to build a typical table that presents the quantitative
comparison of values.

I haven't been able to figure out how to get access to those values. Also, I
have seen there are several packages capable of computing ROC stats, but I
haven't found what I want from any of them. I'm happy to use another one if
that is what you are familiar with.


View this message in context: http://r.789695.n4.nabble.com/Extracting-information-from-ROC-tp4689693.html
Sent from the R help mailing list archive at Nabble.com.

More information about the R-help mailing list