Grant: Discovery Grants Program - Individual

Correlated data arise in many forms in genetic and biomedical problems, such as from studies of pairs of eyes, repeated asthma attacks, or families carrying a disease gene. In order to better understand the disease process we are studying, we need to develop new statistical models that can handle these relationships in our data. The recent progress in genetics, genomics, and other 'omics' research has created new types of correlated data that cannot be analyzed using the current methods. Two important features of these new types of data include how the patients or families were picked as well as how strong or weak the similarity is within families or pairs of eyes or asthma episodes. A statistical method that can handle this type of correlation or similarity is called a frailty model. It can be used to understand and control different types of correlation, such as a common gene, family relationships or specific environmental effects like particular foods or supplements. My research plan is to develop a general statistical framework for handling correlated data from complex study designs based on frailty modeling approach for outcomes that happen over time. Some examples where my research would help are to provide analytical strategies for modeling family-based survival data from complex study design and to provide major public health implication for the genetic counseling of individuals at risk in families carrying disease-susceptibility genes. I will achieve my research plan by working through three goals that will build upon each other: (1) develop a general approach for modeling multivariate survival data with complex correlation structure, (2) take into account how the patients or families were chosen (e.g., from a colon cancer clinic for families with many relatives with colon cancer or from patients who continue to have many episodes of asthma) and (3) adapt my new approaches for specific studies measuring thousands of genes in each person to try to understand how the genes work together. This research program will have many important biomedical applications including risk modeling using family data and estimating the effects of collections of genes from genomic experiments.