Related Experiment Videos
Gamma regression improves Haseman-Elston and variance components linkage analysis for sib-pairs
Mathew J Barber1, Heather J Cordell, Alex J MacGregor
1Department of Medical Genetics, University of Cambridge, Cambridge, UK. mathew.barber@cimr.cam.ac.uk
Genetic Epidemiology
|January 30, 2004
Summary
A new generalized linear model using the gamma distribution improves quantitative trait linkage analysis for sib-pairs. This method overcomes limitations of ordinary least squares and variance components, offering greater robustness and diagnostic capabilities.
Area of Science:
- Genetics
- Statistical Genetics
- Bioinformatics
Background:
- Standard linkage analysis methods for quantitative traits rely on ordinary least squares or phenotypic normality assumptions.
- These methods have limitations in specifying error distributions, leading to misspecification or constraints on the residual coefficient of variation.
Purpose of the Study:
- To address limitations in existing quantitative trait linkage analysis methods.
- To introduce a robust and flexible approach for analyzing sib-pair data.
Main Methods:
- Utilized a generalized linear model (GLM) based on the gamma distribution for linkage analysis.
- Applied the GLM to a sample of unselected sib-pairs.
- The GLM approach was compared to ordinary least squares and variance components methods.
Main Results:
- The gamma-distribution-based GLM overcomes limitations of ordinary least squares by correctly specifying the mean-dependent error distribution.
- It emulates variance components under multivariate normality assumptions but is robust to deviations from normality.
- The GLM provides valuable model-fit diagnostics for linkage analysis.
Conclusions:
- The generalized linear model offers a powerful and robust alternative for quantitative trait linkage analysis in sib-pair studies.
- This approach enhances the reliability of genetic linkage findings by accommodating non-normal distributions and providing model diagnostics.