R 모델 공식
lmer(measurement ~ 1 + (1 | subject) + (1 | site), mydata)
모델에 맞는
Yijk=β0+ηi+θj+εijk
여기서 있는 '번째 행 에서 , 대상인 랜덤 효과 사이트 인 랜덤 효과와 잔여 오차이다. 이러한 임의의 효과에는 모형에 의해 추정되는 이 있습니다. (제목이 사이트 내에 중첩 된 경우 일반적 으로 대신 를 작성합니다 ).Yijkkmeasurement
subject
isite
jηiiθjjεijkσ2η,σ2θ,σ2εθijθj
ICC 계산 방법에 관한 첫 번째 질문에 답하기 위해 : 이 모델에서 ICC는 해당 차단 계수로 설명 된 전체 변동의 비율입니다. 특히 동일한 주제에 대해 무작위로 선택된 두 관측치 간의 상관 관계는 다음과 같습니다.
ICC(Subject)=σ2ησ2η+σ2θ+σ2ε
The correlation between two randomly selected observations from the same site is:
ICC(Site)=σ2θσ2η+σ2θ+σ2ε
The correlation between two randomly selected observations on the same individual, and at the same site (the so-called interaction ICC) is:
ICC(Subject/Site Interaction)=σ2η+σ2θσ2η+σ2θ+σ2ε
It seems you were confused by this being referred to as an "interaction" since it's the sum of individual terms. It's an "interaction" in the sense that it estimates the ICC corresponding to the blocking factor composed on the combination of Subject
and site
- it's important to note that you do not have to include some kind of "interaction" term between the factors to estimate this quantity.
Each of these quantities can be estimated by plugging in the estimates of these variances that come out of the model fitting.
Regarding your second question - as you can see here, each ICC has a fairly clear interpretation. I would argue that the interaction ICC does tell us something interesting - how "similar" are measurements that share both subject and site?
One important point to note is that if subjects are nested within sites, then the Subject
ICC is not meaningful in it's own right, since it's impossible to share Subject
and not site
. Then σ2η becomes only a measure of how much more similar individuals are to themselves, compared to other individuals at their site
.