[R] suggestion for proportions

[csrabak]

Em 9/9/2011 14:32, array chip escreveu:
> Thanks all again for the suggestions. I agree with Wolfgang that
> mcnemar.test() is what I am looking for. The accuracy is the
> proportion of correct diagnosis compared to a gold standard, and I am
> interested in which diagnosis test is better, not particular
> interested in assessing the agreement between the 2 tests.

John,

Since you mention you already have the comparison of each test against 
the golden standard, perhaps the measures obtainable from the confusion 
matrix for especifity, sensitivity, etc. won't help you to ascertain in 
a more complete way?

If you're interested, get a look at epi.tests() in package epiR or 
diagnosis() in package DiagnosisMed.

--
Cesar Rabak