Bayesian analysis of the linear reaction norm model with unknown covariates

Area of Science:

• Quantitative Genetics
• Statistical Modeling
• Evolutionary Biology

Background:

• Reaction norm models are increasingly used to analyze genotype x environment (GxE) interactions.
• Classical models define environmental values using the mean performance of all genotypes, which is often unknown.
• Approximations using phenotypic means can lead to inaccurate genetic inferences.

Purpose of the Study:

• To present a novel Bayesian method for simultaneously inferring environmental values within reaction norm models.
• To evaluate the frequentist properties and accuracy of the proposed method through simulation.

Main Methods:

• Developed a Bayesian Markov Chain Monte Carlo (MCMC) implementation for inferring environmental values.
• Conducted a simulation study to assess the method's performance against true environmental values and phenotypic approximations.
• Compared genetic parameter inferences from the proposed method with those from traditional approaches.

Main Results:

• The proposed Bayesian method accurately estimates model parameters, aligning well with true values in simulations.
• Inferences on genetic parameters using the new method closely match those derived from models with known environmental values.
• Using phenotypic means as proxies for environmental values resulted in significantly poorer inferences.

Conclusions:

• The novel Bayesian approach effectively infers environmental values, enhancing the analysis of genotype x environment interactions.
• This method offers a more robust alternative to approximations using phenotypic means, leading to reliable genetic parameter estimation.
• Accurate environmental value estimation is crucial for precise GxE interaction studies.