[R] Sample size AUC for ROC curves

[Karl Knoblick]

Thanks, but

1) as input for the sample size estimation ony an AUC is given - and the output 
of the study should be an AUC, too. So I thought this should be the right way.

2) I read e.g. in PASS they are doing a sample size calculation for AUC. Are 
thesy wrong?

Sorry for asking further more but I am confused a little....

Karl 



----- Ursprüngliche Mail ----
Von: David Winsemius <dwinsemius at comcast.net>
An: Karl Knoblick <karlknoblich at yahoo.de>
CC: Greg Snow <Greg.Snow at imail.org>; "r-help at stat.math.ethz.ch" 
<r-help at stat.math.ethz.ch>
Gesendet: Samstag, den 13. August 2011, 2:18:37 Uhr
Betreff: Re: [R] Sample size AUC for ROC curves


On Aug 11, 2011, at 5:50 AM, Karl Knoblick wrote:

> Thanks. Actually I thought of something like
> Hanley JA, McNeil BJ. A method of comparing the areas under receiver operating
> characteristic curves derived from the same cases. Radiology. 1983; 148:
> 839–843.
> http://radiology.rsna.org/content/148/3/839.full.pdf+html
> 
> Has anybody R-code for this or something similar but newer?
> 
> The question is just easy - How many subjects do I need if I want to show that
> my diagnostic test is not only a game of dice. Data for input are the epected
> AUC, alpha and beta,....

If you want the binomial choice situation then the AUC is not the right place to 
start. You should be looking at sample size calculations for logistic regression 
(or maybe even binom.test if you have no covariates that matter.)

--David Winsemius

--
> 
> Would be great if somebody has a solution!
> 
> Karl
> 
> 
> 
> ----- Ursprüngliche Mail ----
> Von: Greg Snow <Greg.Snow at imail.org>
> An: Karl Knoblick <karlknoblich at yahoo.de>; "r-help at stat.math.ethz.ch"
> <r-help at stat.math.ethz.ch>
> Gesendet: Dienstag, den 9. August 2011, 19:45:12 Uhr
> Betreff: RE: [R] Sample size AUC for ROC curves
> 
> If you know how to generate random data that represents your null hypothesis
> (chance, auc=0.5) and how to do your analysis, then you can do this by
> simulation, simulate a dataset at a given sample size, analyze it, repeat a
> bunch of times and see if that sample size is about the right size.  If not, 
do
> it again with a different sample size until you find one that works for you.
> 
> --Gregory (Greg) L. Snow Ph.D.
> Statistical Data Center
> Intermountain Healthcare
> greg.snow at imail.org
> 801.408.8111
> 
> 
>> -----Original Message-----
>> From: r-help-bounces at r-project.org [mailto:r-help-bounces at r-
>> project.org] On Behalf Of Karl Knoblick
>> Sent: Monday, August 08, 2011 3:29 PM
>> To: r-help at stat.math.ethz.ch
>> Subject: [R] Sample size AUC for ROC curves
>> 
>> Hallo!
>> 
>> Does anybody know a way to calculate the sample size for comparing AUC
>> of ROC
>> curves against 'by chance' with AUC=0.5 (and/or against anothe AUC)?
>> 
>> Thanks!
>> Karl
>> 
>> ______________________________________________
>> R-help at r-project.org 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.
> 
> 
> ______________________________________________
> R-help at r-project.org 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.