Reanalysis of CVD Risk Factors via Likelihood Methods
We will reanalyze data on risk factors for cardiovascular disease (CVD)including total cholesterol and HDL cholesterol for the subjects of theNHLBI twin study for examinations 1, 2, and 3 of the study. Theanalyses will utilize maximum likelihood estimators of genetic variancewhich are asymptotically more efficient than the method-of-momentsestimators used in previous analyses. The models used will incorporateterms to partition the variance in a trait from twin data into either i)additive genetic variance and unshared environmental variance (the AEmodel), ii) additive genetic variance, dominance genetic variance, andunshared environmental variance (the ADE model), or iii) additivegenetic variance, shared environmental variance, and unsharedenvironmental variance (the ACE model). The AE, ADE, and ACE models canbe fitted separately to data from each of the three exams to obtain across-sectional analysis. We shall also extend these models for usewith longitudinal data by incorporating terms to represent thecovariance of variance components from different exams. The results ofthese longitudinal analyses will yield new insights on genetic effectsaffecting CVD risk factors during the aging process.Two important additional objectives of this proposal are i) to introduceresistant estimation techniques in twin modeling, which trim the effectof outlier data points smoothly, and ii) to carefully study theperformance of maximum likelihood and method-of-moments estimators whenassumptions of the twin model are violated. The results of these partsof the proposal will yield a more complete understanding of the relativemerits and limitations of twin modeling procedures.