Bayes factor for testing the genetic background of quantitative threshold traits

Area of Science:

• Quantitative Genetics
• Statistical Genetics
• Bioinformatics

Background:

• Bayes factor methodology has advanced the evaluation of genetic components in continuous traits.
• Existing methods reparameterize variance component models using intra-class correlation.
• Comparing threshold models with bounded variables requires specialized statistical approaches.

Purpose of the Study:

• To modify the Bayes factor methodology for comparing nested threshold models.
• To develop a robust procedure for analyzing threshold models that differ in a bounded variable.
• To enable the evaluation of genetic influences on traits best described by threshold models.

Main Methods:

• Introduced a data-augmentation step to sample liability, an underlying continuous variable.
• Developed a procedure for sampling from a truncated multivariate normal distribution with non-null covariances.
• Utilized Cholesky factorization and linear restrictions for sampling liability.

Main Results:

• The modified Bayes factor methodology successfully compared threshold models.
• The data-augmentation and sampling procedure yielded satisfactory results in computer simulations.
• The approach effectively handles the complexities of threshold models and bounded variables.

Conclusions:

• The presented modification extends Bayes factor analysis to threshold models.
• This method provides a valuable tool for assessing genetic contributions to categorical or binary traits.
• The procedure is computationally sound and validated through simulations.