SO WEIRD!! About degrees of freedom

cxf

New Member
#1
Hi,

I always thought the degrees of freedom should be a whole number. However, I was told by a faculty member today that the df value could be fractional, especially in independent t-tests.

I am wondering if anybody has seen a fractional df value???!!!

Thanks!!!
 

mp83

TS Contributor
#2
Yes, I've seen them, there are out there:)

They are used in applications, e.g when "correcting" the df's in the F-test of a repeated measures analysis [Greenhouse and Geisser (1959)]