CMStatistics 2018: Start Registration
View Submission - CMStatistics
Title: Median-based multifactor dimensionality reduction methods for the survival phenotype Authors:  Seungyeoun Lee - Sejong University (Korea, South) [presenting]
Abstract: With the development of high throughput technologies for genetic variants, genome-wide association studies on complex diseases such as hypotension, diabetes and cancers have been extensively developed for the last decades. The multifactor dimensionality reduction method has been originally proposed to reduce high order dimensions in gene-gene interaction analysis, in which the high-level genetic variants are classified into high and low risk groups by the ratio of the cases and controls for a case-control study. Many modifications for the multifactor dimensionality reduction methods have been proposed by allowing the various classifiers and generalizing the phenotypes. The quantitative multifactor dimensionality reduction method uses a $t$-test statistic to classify the high and low risk groups for the continuous phenotype. We propose to use a median survival time as a classifier for the survival phenotype, which called the median-based MDR. We compare the power of the proposed median-based MDR with the Surv-MDR in the simulation study and analyze a real example of the ovarian cancer patient data.