How to calculate reliability of a change score (SAS)

#1
Hi all,

I'm hoping you can lend me your expertise on this somewhat obscure problem. I have some repeated measures data (time 1 measurement, time 2 measurement, change/difference score). I'm interested in calculating the reliability of the change/difference score.

Traditional scale construction statistics (i.e., internal consistency) don't appear work because the measure isn't a composite of different scale items.

It also doesn't work to use inter-rater reliability because the measure isn't being made subjectively by people. Rather, it is the same objective measurement taken at 2 points in time with the difference of the 2 measurements being the variable I'm interested in examining for reliability.

I would really appreciate any thoughts you might have about the best way to calculate the reliability of this variable...or if you think it is even possible. If it helps, SAS is the statistical package of choice, but general guidance is also very appreciated.

Thank you all so much!
 

Karabiner

TS Contributor
#2
If the same shift was supposed for all subjects, regardless of baseline level, then the correlation between t1 and t2 values can be used to assess the reliabilty of the measurement. If you have the reliabilty of the measurement, then you can calculate the reliabilty of its change score. But maybe the first part of my reasoning is wrong.

With kind regards

K.