A note on the accuracy of PAC-likelihood inference with microsatellite data

Area of Science:

• Population Genetics
• Computational Biology
• Statistical Genetics

Background:

• Importance sampling is a common method for likelihood computation in population genetics.
• Existing methods can be computationally intensive, especially for large datasets.
• Approximating conditional probabilities offers a potential simplification.

Purpose of the Study:

• To evaluate the accuracy and efficiency of the product of approximate conditionals (PAC) likelihood method.
• To assess the performance of PAC likelihood with microsatellite data.
• To compare PAC likelihood with traditional importance sampling (IS) algorithms.

Main Methods:

• Utilized simulated microsatellite data for analysis.
• Applied the product of approximate conditionals (PAC) likelihood approximation.
• Compared PAC likelihood results with those obtained from importance sampling (IS).

Main Results:

• PAC likelihood demonstrated a negligible bias across a wide range of the scaled mutation parameter (theta).
• PAC likelihood exhibited lower sampling variance compared to IS.
• PAC likelihood resulted in reduced computation time relative to IS.

Conclusions:

• The PAC likelihood method is a highly accurate and efficient alternative to IS algorithms for population genetics.
• PAC likelihood is suitable for computer-intensive inference methods like MCMC.
• This method offers a valuable tool for analyzing population genetic data, particularly microsatellites.