The article by Peter Beerli compares the performance and accuracy of Bayesian and maximum-likelihood (ML) inference methods for estimating population genetic parameters using the program MIGRATE. Both methods use the same Markov chain Monte Carlo (MCMC) algorithm but differ in their parameter proposal distributions and likelihood function maximization. Using simulated datasets, the Bayesian method generally outperforms the ML approach in terms of accuracy and coverage, although both methods can be equally effective under certain conditions. The Bayesian framework, with appropriate prior distributions, addresses some of the limitations of the ML approach, such as the difficulty in handling sparse data and the delivery of non-conservative support intervals. The program MIGRATE has been extended to support both ML and Bayesian inference, allowing for direct comparison of the two methods. The results show that the Bayesian approach provides more reliable estimates, especially for low-variability datasets, while the ML approach often fails to converge and delivers less accurate results. The conclusion suggests that the Bayesian approach is preferable for users interested in parameter estimates and their support intervals, particularly when dealing with low-information data.The article by Peter Beerli compares the performance and accuracy of Bayesian and maximum-likelihood (ML) inference methods for estimating population genetic parameters using the program MIGRATE. Both methods use the same Markov chain Monte Carlo (MCMC) algorithm but differ in their parameter proposal distributions and likelihood function maximization. Using simulated datasets, the Bayesian method generally outperforms the ML approach in terms of accuracy and coverage, although both methods can be equally effective under certain conditions. The Bayesian framework, with appropriate prior distributions, addresses some of the limitations of the ML approach, such as the difficulty in handling sparse data and the delivery of non-conservative support intervals. The program MIGRATE has been extended to support both ML and Bayesian inference, allowing for direct comparison of the two methods. The results show that the Bayesian approach provides more reliable estimates, especially for low-variability datasets, while the ML approach often fails to converge and delivers less accurate results. The conclusion suggests that the Bayesian approach is preferable for users interested in parameter estimates and their support intervals, particularly when dealing with low-information data.