Hierarchical Linear Models and Structural Equation Modelling for the Children of Siblings model

The main aim of this paper is to examine and apply the Children of Siblings model to normally distributed outcome data. The model refers to siblings who are mothers/aunts and/or fathers/uncles and on whose children an outcome variable is measured. This is done through two statistical models. First a Hierarchical Linear Model (also known as multi-level model) which takes into account clustering of data in three levels. The levels correspond to increasing sizes of a family: individual, nuclear family and extended family. This model makes it possible to examine whether a specific exposure is responsible for a change in the outcome variable after confounders and clustering have been considered. Furthermore the genetical relatedness is taken into account making it possible to draw conclusions regarding the familial effects on the association between exposure and outcome. The second model is a 2-level Structional Equation Model which aims at splitting up variance/covariance in three different predetermined components representing genetic influence, environmental similarity and environmental dissimilarity using two levels of clustering in the data. The clustering levels are withinand betweennuclear families, which also are considered hierarchical. This model investigates the importance of genetic heritability contra environmental heritability. An example study is performed using data for smoking habits of mothers during pregnancy and the ”psychological functioning capacity”, a prognosis of the ability to cope with war-time stress, of their children as measured by Försvarsverket during the medical examination for military service. We found that smoking during pregnancy does not affect the psychological functioning capacity. The exposures effect on the outcome has familial confounding, and this is mainly explained through genetical relatedness