Intraclass Correlation, Median Odds Ratio, and Ecological Reliability

Abstract

The part of the variability on the neighborhood and the school level is not visible when looking at both levels separately. This is only possible with models that use cross-classified data. For these models, the intraclass correlation (ICC), the median odds ratio, and the ecological reliability (λ) were analyzed. If neighborhood and school units are taken into account simultaneously at the aggregate level, the variance of the dependent variable and the ICC for the neighborhoods are reduced drastically or completely disappeared. The ICC between schools remains stable or even increases. It turns out that the schools as aggregate units have a higher potential for the explanation of self-control and juvenile delinquency than the neighborhoods.