Statistical Linkage Across High-Dimensional Observational Domains

• Activity Number: 183
• Type: Contributed
• Date/Time: Monday, August 5, 2013 : 10:30 AM to 12:20 PM
• Sponsor: ENAR
• Author(s): Leonard Hearne*+ and Toni Kazic
• Companies: and University of Missouri at Columbia
• Keywords: Genotypic Analysis ; Phenotypic Analysis ; Nonparametric Association ; Regional Associations ; High Dimensional Data

Abstract

It is common in many experimental sciences to have high-dimensional data from multiple sources collected on the same experimental units. For example, one study may include data whose domains are genotypic and phenotypic. When we are interested in comparing relationships between homogeneous regions in one high dimensional domain with one or more regions in another high dimensional domain, the number of possible comparisons of interest may be extremely large and not known a priori.

We outline procedures for identifying possible relationships between regions in one domain and regions in another domain. These procedures are particularly useful when the number of relationships is unknown, and when the relationships may be of unknown complexity. The associations between domains may be one to one through many to many. If the data are dense enough, then statistical measures of association can be estimated. These procedures have the capacity to identify and measure the probability of inter-observational domain associations of mixed complexity.