Hi,
A couple of years ago I used data from Olympic figure skating judges to
learn generalizability theory. It's a way of partitioning variance when
you have multiple raters rating multiple targets on multiple dimensions.
You partition the variance into that attributable to the raters (judges),
targets (skaters), and dimensions (technical vs. artistic). You can then
derive measures of reliability and agreement. You can also check to see if
the reliability increases by deleting any of the judges. As I recall, the
scores were pretty reliable, but the surprising part had to do with the
reliability of the technical aspects vs. the artistic aspects. I expected
greater agreement on the technical (more objective?) scores; but I found
greater agreement on the artistic scores. This held across multiple
competitions (men's, women's, and pairs). It's a strange "sport" alright.

Al

Al B. Shealy
Columbia State U.

---
You are currently subscribed to tips as: [EMAIL PROTECTED]
To unsubscribe send a blank email to [EMAIL PROTECTED]

Reply via email to