The study looked at different models for analyzing data from repeated measurements, focusing on how different transformations of the data affect the results. By using a concept called parameter orthogonality, the researchers found that the intraclass correlation coefficient remains consistent regardless of the transformation used. This finding applies to various types of models, including those with complex variance components and correlation structures. The results suggest that it may be beneficial to describe the covariance structure in these models using separate parameters for variance and correlation.