Analysis of familial aggregation using recurrence risk for complex survey data Open Access
Downloadable ContentDownload PDF
Familial aggregation of a disease is important for studying possible genetic etiologyof a disease. A popular and useful measure of family aggregation is recurrence risk.It is the probability that a family member of an affected individual is also affected.A problem when estimating recurrence risk is how to allow for varying the familysizes. The quadratic exponential model (QEM) has been proposed for estimatingparameters for computing the recurrence risk when there is varying family sizes inthe sample. In this thesis, we use a marginal estimation approach for the QEMbased on a pairwise composite likelihood to address varying family sizes. Householdhealth surveys with (family) network sampling, which surveyed individuals reportabout disease status of themselves and specified relatives, has been shown to be useful for estimating prevalence of diseases and more recently for estimating recurrencerisk of disease using nonparametric classical survey methods. Because these surveyshave complex sample designs such as stratied multistage cluster designs with sample weighting for differential sample selection rates, this thesis extends the compositepairwise likelihood estimation and hypothesis of parameters of the QEM for simplerandom samples to data from these complex sample designs. In addition, the QEMis extended to simultaneously estimate and test parameters and recurrence risk formultiple family relationships, for comparing recurrence risk across family-level covariates (e.g., race) and utilizing propensity score weighting to adjust for confounding byindividual-level covariates (e.g., age). Simulations are used to study the nite sampleproperties of the parameter estimation, variance estimation and level and power ofhypotheses testing based on derived Wald and Quasi-score tests for these extendedQEMs. Finally, our methods are illustrated using the 1976 National Health InterviewSurvey diabetes data set.