Shi et al. [2009] applied a three-level hierarchical linear mixed-effects model for testing genetic main effects and gene×age interactions affecting coronary artery calcification, while accounting for correlation due to the family-based longitudinal data. Genome-wide association analyses using the 50k chip were conducted based on cross-sectional data (i.e., each of the three single visit data sets separately) and also on the longitudinal data (i.e., using data from all three visits simultaneously). They had prior knowledge of the simulation schema and answers. Results showed that the association tests using longitudinal data were more powerful than those using cross-sectional data. Out of the five simulated major gene SNPs of coronary artery calcification, association with rs17714718 (τ2) was detected only when using the longitudinal data. SNP rs213952 (τ5) was found to be significant with both longitudinal and cross-sectional data, but the former yielded a more significant result. None of the other major gene SNPs were found to be significant. No significant gene×age interactions were observed.
