Addressing intraclass correlation coefficient issues


New Member
ICC estimates changes according to the magnitude of the data. If the data are >1, ICC values are going to be way closer to 1 than if the data are decimals. For example, if you are calculating ICC between three projections of 2010's % growth in each sector, the ICC may be 0.95 if these three projections are similar. However, if you are calculating the ICC between the differences between the projections and the actual growth, (the difference we assume is very small), then the ICC value will probably be way lower.

Is there a way to address this issue?