Bayesian synthetic likelihood (BSL, Price et al (2018)) is an alternative to standard, non-parametric approximate Bayesian computation (ABC). BSL assumes a multivariate normal distribution for the summary statistic likelihood and it is suitable when the distribution of the model summary statistics is sufficiently regular.
In this package, a Metropolis Hastings Markov chain Monte Carlo (MH-MCMC) implementation of BSL is available. We also include an implementation of BSLasso (An et al., 2018) which is computationally more efficient when the dimension of the summary statistic is high.
Parallel computing is supported through the foreach package and users can specify their own parallel
backend by using packages like doParallel or doMC.
The main functionality is available through
bsl: The general function to perform BSL or BSLasso (with or without parallel computing).
selectPenalty: A function to select the penalty for BSLasso.
Several examples have also been included. These examples can be used to reproduce the results of An et al. (2018).
ma2: The MA(2) example from An et al. (2018).
mgnk: The multivariate G&K example from An et al. (2018).
cell: The cell biology example from Price et al. (2018) and An et al. (2018).
Extensions to this package are planned.
An, Z., South, L. F., Nott, D. J. & Drovandi, C. C. (2018). Accelerating Bayesian synthetic likelihood with the graphical lasso. https://eprints.qut.edu.au/102263/
Price, L. F., Drovandi, C. C., Lee, A., & Nott, D. J. (2018). Bayesian synthetic likelihood. To appear in Journal of Computational and Graphical Statistics. https://eprints.qut.edu.au/92795/