1
$\begingroup$

I'm interested in learning more about the Q value produced by val.prob in the R package RMS and how it compares with the Brier score. I understand Q, the quality index, is a variation of the logarithmic scoring rule that can be decomposed in useful ways, but I haven't been able to find guidance on:

--What values of Q are acceptable? --What is the range of Q? --How is Q related to the simple logarithmic scoring rule? Does the translation of the logarithmic score into Q make for easier interpretation? --In what cases is the Brier score or Q/logarithmic score preferred, and why might the conclusions these scores lead to differ?

I haven't been able to find great references to the Q score or discussions of the different aspects of accuracy measured by different scoring rules -- if you can provide such references those would be great to read too.

$\endgroup$
1
$\begingroup$

Judging the magnitude of Q is not so easy. But more information may be obtained at http://biostat.mc.vanderbilt.edu/wiki/pub/Main/FrankHarrell/logistCal.pdf

The source code for val.prob gives the formula for Q but look at the above paper first.

| cite | improve this answer | |
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.