I rarely read of it's use in papers - is it problematic/does it complicate the picture?
.
it is of course desirable to have just one aggregate combining both error rate and time to respond ...
but ...
being an aggregate, it may of course hide some fine and relevant structure in the interaction between its components
(the same could be said for F-score in information retrieval)
and do **not** use IES if you suspect some possible accuracy/speed trade-off in your experiments !
see for instance :
http://www.psychologicabelgica.com/article/view/pb-51-1-5
.
In my opinion, the IES is a rather violent treatment of RT and accuracy, and obviously problematic in the case of speed-error tradeoff. A much more differentiated approach is survival analysis, where the likelihood of an error is modelled on the basis of response speed. Panis & Hermens (2014) wrote a very cool paper on what they call Event History Analysis of error rates and reaction times.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics