I would like to determine one ICC (interclass corelation coeficient) for method A and method B. I have measurements of 10 objects at several characteristic places. Is it correct to combine all these measurements for a common ICC analysis?
Should I determine several ICC for each of these measurement locations separately?
Hello Pawel!
I think the answer to this question relies heavily on what your research objective is. If you want to investigate whether method A and method B yield similar results regardless of the conditions of the measurement (i.e. the place), then you should combine all the measurements in a common ICC analysis. Combining all measurements should also increase the statistical power of this experiment.
If, on the other hand, you are looking for particular conditions under which A and B cease to agree, then calculating ICCs separately is the way to go. Using confidence intervals, you could actually prove that there exist particular places where A and B are significantly more likely to deviate. Unfortunately it is not possible to judge whether A or B is incorrect, at least not without access to some sort of "ground truth".
My research goal is to compare the two methods in general, but I am not sure if the approach of combining several measurement points from one object is correct.
- Badges
- Science topic
- Similar topics
- Social Psychology
- Attitude