Best of both worlds: Psychometric Analyses of Latent and Observed Genetically Informative Data

The goal of Chapter 2 was to introduce structural equation modeling (SEM) in genetics. I explained how familial relationships are used to infer unmeasured genetic and environmental effects, reviewed popular genetic SEM models, and highlighted recent developments. Initially, SEM in genetics focused on decomposing phenotypic variance into additive genetic (A), shared environmental (C), dominance genetic (D), and unique environmental (E) components. Recent advancements include models incorporating genetic-environmental interactions, correlated effects, and measured genetic variables. In Chapter 3, I compared two methods for detecting passive genotype-environment covariance due to genetic and cultural transmission. The nuclear twin family (NTF) design uses phenotypic data from twins and parents, while the transmitted-nontransmitted (T-NT) alleles design uses genotype data from parents and offspring. The NTF design showed greater power for detecting cultural transmission. A third approach added polygenic scores (PGSs) to the NTF design but showed no significant improvement due to the limited predictive power of current PGSs. Future increases in PGS explanatory power could enhance detection in both designs. Chapter 4 examined intelligence and its relationship with psychopathology measures—negative affect, anxiety, ODD, autism, and ADHD—using data from Dutch twin pairs. Negative correlations between intelligence and these traits ranged from -0.09 (negative affect) to -0.14 (ADHD). Bivariate ACE modeling revealed pleiotropic genetic effects primarily explained the negative correlations for anxiety, autism, and ADHD. Additionally, Purcell moderation modeling showed intelligence moderated genetic effects on negative affect and environmental effects on anxiety. In Chapter 5, I applied the environment-by-PGS model to negative affect and anxiety using IQ PGSs. This model estimated genotype-environment interaction based on twin designs. The IQ PGS moderated shared environmental effects on negative affect, with heritability highest in children with below-average intelligence. The model estimates additive genetic x shared environment (AxC) and additive genetic x unique environment (AxE) interactions, leveraging the Purcell framework with PGSs as moderators. Although PGSs currently capture only part of additive genetic variance, simulations demonstrated this approach’s feasibility under realistic conditions. The method could be extended to compare interactions across SES or age groups. Chapter 6 explored genetic and environmental contributions to aggression across the lifespan. Using IRT modeling, I tested measurement invariance for aggression data collected across age groups (3–90 years). The data showed invariance and allowed fitting an autoregressive simplex model to latent genetic and environmental variables. Phenotypic correlations ranged from 0.34 at age 3 to 0.68 at ages 50–59. Heritability was highest in adolescence (0.74 at age 12) and decreased with age, while unique environmental variance increased through early adulthood. Genetic factors drove most stability in aggression, with evidence of genetic innovations before age 25 but none thereafter. Chapter 7 summarized the contributions of the Netherlands Twin Register to the Genetics of Early Milestones and Skills (GEMS) consortium. Genome-wide association studies (GWASs) on four dimensions of infant temperament in two-year-olds were described, including details of phenotype and genotype data, quality control, and results.