Related Experiment Videos
Search for faster methods of fitting the regressive models to quantitative traits
F M Demenais1, C Murigande, G E Bonney
1Division of Biostatistics and Epidemiology, Howard University Cancer Center, Washington, D.C. 20060.
Genetic Epidemiology
|January 1, 1990
Summary
Regressive models analyze familial trait patterns. Class A models risk false major gene detection, but careful testing reduces this. Class D models benefit from approximations for faster computation.
Area of Science:
- Quantitative genetics
- Statistical genetics
- Biostatistics
Background:
- Regressive models analyze familial dependence of quantitative traits using regression relationships between phenotypes and genotypes of individuals and their antecedents.
- Complex familial patterns, particularly with multiple siblings (Class D models), pose computational challenges for likelihood calculation as the Elston-Stewart algorithm is not generally applicable.
- Simpler Class A models, while computationally efficient, may incorrectly infer a major gene due to restrictions on sibling correlations.
Purpose of the Study:
- To investigate the robustness of the Class A regressive model against the false inference of a major gene.
- To identify faster computational methods for likelihood calculation in Class D regressive models.
Main Methods:
- A simulation study was conducted to assess the Class A model's susceptibility to spurious major gene detection.
- Various approximations of the likelihood formulation for Class D models were evaluated.
- Tests for Mendelian transmission and absence of major gene effects were employed to refine Class A model analysis.
Main Results:
- The Class A model's robustness is compromised by sibling correlations exceeding model specifications; careful transmission probability testing significantly reduces false major gene detection (from 26-30 to 0-4 out of 30 replicates).
- Approximations 6 and 8 for the Class D model's likelihood formulation demonstrated appropriate performance in parameter estimation and hypothesis testing across different generating models.
- These approximations enable the use of the Elston-Stewart algorithm, substantially reducing computation time for Class D models.
Conclusions:
- Careful application of transmission probability tests is crucial for accurate major gene inference with Class A regressive models.
- Selected approximations provide efficient and accurate likelihood computation for Class D regressive models, facilitating complex familial data analysis.
- The findings offer improved methodologies for both robust genetic analysis and computational efficiency in quantitative genetics.