A Comparison of Statistical Approaches for Detecting Genes Influencing Poisson-Distributed Traits

by Mark J. Giganti, Nathan A. Johnson, and Steven T. Garren.

Abstract: Abstract: In the field of statistical genetics, many methods for identifying quantitative trait loci (QTLs), which are genes influencing phenotypes, require the assumption of normality in the phenotypic distribution. However, many applications of QTL analysis do not have normally distributed phenotypes. In this paper, we compare different parametric and nonparametric approaches for detecting the presence of QTLs when phenotypes follow a Poisson distribution. Using parametric tests, a Wilcoxon rank-sum test, and permutation methods, we analyzed simulated genotypes and phenotypes of mice. Our simulations suggest that the parametric tests are as effective as the nonparametric tests and computationally faster than the permutation tests. We conclude that parametric approaches are capable of performing QTL analysis for Poisson distributions, even for small phenotypic means.

Key Words: Backcross data, Permutation tests, Poisson distribution, Quantitative trait loci

Authors:
Mark J. Giganti, giganti00@yahoo.com
Nathan A. Johnson, johnsonna@wlu.edu
Steven T. Garren, garrenst@jmu.edu

Editor: Chris Amos,camos@request.mdacc.tmc.edu.sg

READING THE ARTICLE: You can read the article in portable document (.pdf) format (188144 bytes.)

NOTE: The content of this article is the intellectual property of the authors, who retains all rights to future publication.

This page has been accessed 4362 times since July 24, 2006.


Return to the InterStat Home Page.