Reanalysis of CVD Risk Factors Via Likelihood Methods

BACKGROUND:

The results of these longitudinal analyses yielded new insights on genetic effects affecting CVD risk factors during the aging process.

DESIGN NARRATIVE:

The analyses utilized maximum likelihood estimators of genetic variance which were asymptotically more efficient than the method-of-moments estimators used in previous analyses. The models used incorporated terms to partition the variance in a trait from twin data into either i) additive genetic variance and unshared environmental variance (the AE model), ii) additive genetic variance, dominance genetic variance, and unshared environmental variance (the ADE model), or iii) additive genetic variance, shared environmental variance, and unshared environmental variance (the ACE model). The AE, ADE, and ACE models were fitted separately to data from each of the three exams to obtain a cross-sectional analysis. The investigators also extended these models for use with longitudinal data by incorporating terms to represent the covariance of variance components from different exams.

Two important additional objectives of this study were i) to introduce resistant estimation techniques in twin modeling, which trimmed the effect of outlier data points smoothly, and ii) to carefully study the performance of maximum likelihood and method-of-moments estimators when assumptions of the twin model were violated. The results of these parts of the study should yield a more complete understanding of the relative merits and limitations of twin modeling procedures.