The summary-likelihood method and its implementation in the Infusion package.

Area of Science:

• Population genetics
• Computational biology
• Statistical inference

Background:

• Approximate Bayesian computation (ABC) is widely used for parameter inference in population genetics when likelihood functions are intractable.
• Existing simulation methods like ABC often rely heavily on prior distributions, potentially biasing results.

Purpose of the Study:

• To present a novel likelihood-based approach, termed summary likelihood, for parameter inference in population genetic models.
• To provide an automated implementation of this method in an R package named Infusion.
• To evaluate the performance of summary likelihood, particularly for inferring population size changes from genetic data.

Main Methods:

• Development of the summary likelihood method, which analyzes the information within simulated summary statistics.
• Implementation of the method as the R package 'Infusion'.
• Testing the method on a population genetics scenario involving population size change inference.

Main Results:

• The summary likelihood method yields confidence intervals with controlled coverage, independent of prior distributions.
• This offers an advantage over approximate Bayesian computation, which is sensitive to prior choices.
• The method is demonstrated to be applicable to models with at least six parameters.

Conclusions:

• Summary likelihood provides a robust alternative for parameter inference in population genetics, offering improved reliability over existing simulation-based methods.
• The Infusion R package facilitates the application of this method.
• Further modifications are discussed for extending the method to higher-dimensional inference problems.