Last working group of the semester!
Wednesday, November 2, 2016
Building 2, Room 426, 12:30 – 1:00 pm
Matthew Goodman, Doctoral Student
Department of Biostatistics
Harvard T.H. Chan School of Public Health
Genetic Tests for Alzheimer’s Data using Kernel Regression
Genetic association studies are sometimes nested within longitudinal cohorts which provide a rich source of phenotype data. Taking advantage of these phenotypes often requires the development of new genetic tests. One setting which has not previously received much attention involves ‘zero-inflated’ phenotype data that contains excess zeros compared with standard distributions. For example in Alzheimer’s research this outcome could be the count of neurofibrillary tangles, or the continuous measurement of amyloid burden in postmortem brain tissue pathology studies. Recent research into statistical tests for genome-wide association methods has shown that the kernel regression variance components score test for a group of genomic markers has good statistical power with respect to marginal (single marker) tests in various outcome settings, especially when there is correlation among the markers within a group. Previously this approach has been developed in the generalized linear model and time-to-event data settings. In the first part of this talk I will discuss a new kernel regression score test that can be applied to a Zero-Inflated Poisson outcome. I will also discuss ongoing research on testing genetic associations with a longitudinal cognitive decline phenotype.
