[Oradlist] Help with a word

Ann Wenzel awenzel at odont.au.dk
Sun Nov 26 23:52:05 PST 2006


In the textbooks I know, these terms are defined as follows:

Reliability=accuracy+precision (the same as reproducibility)

A test can be reliable in terms of accuracy - meaning that it actually 
reflects the "truth" (a proper gold standard is needed). Accuracy may be 
described in various ways (true positive fraction (sens), true negative 
fraction (spec) etc.) or ROC areas or similar.

A test can be reliable in terms of reproducibility (or precision) meaning 
that the same result is obtained every time you do the test - (but not 
necessarily the correct result); described by intra- or interobserver 
agreement, usually tested with kappa statistics or the like.

In Sharon's case, the authors seem to evaluate accuracy since they know what 
structures are to be presented in the images.
If someone can quote from other texbooks with different definitions, it 
would be interesting to know.

Ann 



More information about the Oradlist mailing list